... which the paths of task performance and information needed cannot be determined in advance For the former, information need may require task- related information for performing the task For the latter,... relation to sources and channels of information, including both active and passive information seeking, and information use Information seeking behavior is the purposive seeking for information as... Mastery Referent Information Role Clarification Performance Information Social Integration Social Information Acculturation Organizational Information Figure 2-1: Information Types and Information Needs
Acknowledgement I would like to thank many people who have seen me through my research work. Firstly, I greatly appreciate my respectable supervisors Dr. Bernard Tan Cheng Yian and Dr. Xu Yunjie (co-supervisor) for their invaluable guidance, support and encouragement given to me during this project. They had illuminated many of my questions and doubts and constantly provided me with every useful resource relevant to my research topic. Secondly, I would also like to thank my lab-mates who had offered interesting ideas that helped improve the design of my study. Thirdly, I also gratefully acknowledge the financial support by the School of Computing, National University of Singapore (Research Grant: R253-000-028-112). Finally, I am pleased those departments in National University of Singapore that supported this research study. And, I also thank the hundreds of respondents who provided the data for the study that I report here. i Table of Content 1. Introduction .................................................................................................................. 1 2. Conceptual Background .............................................................................................. 6 2.1 Information Needs.....................................................................................................7 2.1.1 Types of Task Attributes ....................................................................................8 2.1.2 Types of Information Sought............................................................................11 2.2 Information Source Preference................................................................................14 2.2.1 Categories of Information Sources ...................................................................15 2.2.2 Factors on Information Source Preference ......................................................16 3. Research Model and Hypotheses Development ....................................................... 27 3.1 Cost-benefit Framework ......................................................................................27 3.2 Elaboration Likelihood Model............................................................................28 3.3 Research Model ..................................................................................................30 4. Research Methodology ............................................................................................... 36 4.1 Item Generation and Content Validity ....................................................................36 4.2 Scale Evaluation......................................................................................................42 4.3 Survey Administration ............................................................................................44 5. Data Analysis and Results.......................................................................................... 46 5.1 Measurement Model ............................................................................................47 5.2 Structural Model ..................................................................................................51 6. Discussion and Implication........................................................................................ 53 6.1 Discussion ...............................................................................................................54 6.2 Implications.............................................................................................................55 7. Limitations and Future Studies .................................................................................57 8. Conclusion ................................................................................................................... 58 References………………………………………………………………………………59 Appendix ii Summary With massive amounts of information generated every day, information overload has become a well recognized issue of today. Not only does information overload affect individuals’ everyday life, but also it affects individuals’ job performance in organizations. In addition, same information can be available from multiple sources which are classified into four categories: personal vs. impersonal and internal vs. external. To effectively gain information needed, it is likely that individuals need to make a choice among various sources. However, previous research either focused on how a seeker chose an information source or on what affected the amount of information seeking, yet paid little attention to the role of task attribute in relation to both benefit- and cost- factors as moderated effect for information source preference. By considering interpersonal source preference from the perspective of information seekers, this thesis proposed a research model based on integrating cost-benefit framework with elaboration likelihood model. Both benefit- and cost- factors were generated including perceived content quality, physical proximity, source understandability and social risk; and task importance was examined as a moderator for information source preference. Furthermore, a field survey study was administrated to 161 knowledge workers in a top-tier university in Singapore who had sought task relevant information in the course of their work. The findings of this study provide preliminary support for the proposed model and indicate that perceived content quality significantly affects seekers’ information source iii preference. Implications of the results for further research and improving information management are then discussed. The findings of this study suggest the least effort principle might not apply to the context of single source type. Our result runs head-on against the least effort principle. However, it is consistent with some other studies which also found quality to be more important than accessibility. Therefore, mangers who want to encourage information and knowledge seeking in an organization, this study suggest that enriching the information environment is an important guideline for job design. Given the assumption of personal information seeking dominant, job design should try to place people of related skills together, so that good quality sources can be accessed with the least effort. Moreover, it is important to make seekers aware of whom the expert is. Furthermore, further studies should incorporate multiple sources and compare seekers’ decisions with and across sources. iv 1. Introduction With the development of information technology, not only does information overload influence individuals’ everyday life, but also it affects information management in organizations. In the USA, for example, there are ten thousand of newspapers and magazines, more than 100,000 new book titles published every day, and over 60 billion pieces of advertising junk mail coming into mail boxes every year. As reported, the world's total yearly production of print, film, optical, and magnetic content would require roughly 1.5 billion gigabytes of storage. This is the equivalent of 250 megabytes per person for each man, woman, and child on earth (Lyman et al., 2003). Moreover, with the growing of a vast array of network services around the Internet, massive amounts of information are increasing added everyday. Not only does large amounts of information take individuals’ time to consume, but also distract individuals’ attention. In organization settings, however, the same knowledge may be stored by various workers in everywhere in spite of corporate intranet, vast network of servers, and plenty of business intelligence tools settled in organizations (Stuart, 2001). On the other hand, information is viewed as vital to effective decision making under the volatile environment of today (e.g., Choo, 2001). Individuals need to seek information to control uncertainty at a certain level for performing job tasks, improving job performance and maintaining comfortable social relationship within organizations (Morrison, 2002). In a longer term, employees actively engaging in information seeking exhibit more positive attitude towards their jobs, themselves, and higher loyalty to the organization (Morrison, 2002). Therefore, with the fact that a large amount of additional information is available to individuals and also the same information may be obtained by various sources, efficient and effective discovery -1- of resource and knowledge has become a well recognized research issue (e.g., Montebello, 1998). Information seeking refers to purposive seeking for information as a consequence of a need to satisfy some goal (Wilson, 2000). Definitely, such behavior is triggered by information needs. In organization settings, information needs can be classified into four categories including task mastery, role clarification, acculturation and social integration (e.g., Morrison, 1993b). Each kind of information needs requires different information types; consequently, a seeker likely needs to use different sources with respect to information needs (e.g., Morrison, 1993a; 1993b). A cost-benefit paradigm is widely used in the information seeking literature. However, there is disagreement on the relative importance of the two components. Two research streams have been identified in information source preference: perceived quality or value (e.g., Ashford, 1986; Morrison & Vancouver, 2000), and the least-effort principle (e.g., Yitzhaki & Hammershlag, 2004; Anderson et al., 2001; Chakrabarti et al., 1983; Culnan, 1983; Gerstberger & Allen 1968). The former emphasizes on perceived quality of information as a dominant factor for information source selection. The latter emphasizes on the least psychological and financial costs of information source selection (e.g., Allen, 1977; Rosenberg, 1967). Prior research on information seeking behavior has reported that personal source was one key source for information seekers (e.g., Leckie et al., 1996; Morrison, 1993a). While cost-benefit model is frequently used to explain a seeker’s information source preference, both or either perceived quality and perceived costs do not ensure the consequences of information source decision. This is due to that information seeking greatly depends on task conditions (e.g., Byström, 2002; Vakkari, 1999; Leckie et al., -2- 1996; Byström & Järvelin, 1995; Belkin et al., 1982). For example, when the issue is perceived important, a seeker is willing to exert much cognitive effort and take long time to assess the issue relevant information obtained. Under such circumstance, information source with high quality of information tends to be chosen. However, when the issue is not perceived important, a seeker may only judge the issue relevant information in terms of source characteristics, such as source credibility (e.g., Petty & Cacioppo, 1986). Thus, to explore the criteria of information source preference, task context should be involved in the study of information seeking. Previous research on information source selection mainly has focused on task complexity (e.g., Byström & Järvelin, 1995; Culnan, 1983; O’Reilly, 1982) and task uncertainty (e.g., O’Reilly, 1982; Kuhlthau, 1999). For simplifying task complexity, a seeker chooses more information sources (e.g., Byström & Järvelin, 1995). For resolving task uncertainty, a seeker chooses various information sources with respect to information types required. Both directions of research emphasize how a seeker views task attribute as a barrier, and consequently handles the barrier by means of information seeking. However, task attribute may also be regarded as a motivator because a seeker possibly obtains some rewards from the consequences of the task such as monetary rewards, promotion, or comfortable social relationship. Under such condition, a seeker may undertake different information source preference. In addition, most previous studies in which researchers have concerned either how does a seeker choose information source across information types (e.g., Kwasitsu, 2003; Choo, 1994) such as personal vs. documentary sources? Or what affect the amount of information a seeker obtain (e.g., Morrison & Vancouver, 2000; Tan & Zhao, 2003)? Yet few researchers study information source preference by within person approach except -3- for the works conducted by Morrison and Vancouver (2000; Vancouver & Morrison, 1995) in which they involves need for achievement as the moderator for information seeking. However, they merely concern the moderation of need for achievement on source expertise but not on source accessibility. In this study, we apply cost-benefit framework to explore how a seeker selects information sources by cost- and benefitfactors trade-off. Besides, we integrates elaboration likelihood model with cost-benefit framework to indicate how the consequences of the task under consideration affect the importance of the costs and benefits of selecting information sources. The purpose of this study is to investigate seekers’ choice of information sources and particularly seekers’ choice of personal sources. Two research questions are proposed as following: 1. What cost-benefit factors affect a seeker choosing personal sources for taskrelated information? 2. How does task importance moderate the impact of cost-benefit factors on information source preference? We focus on interpersonal information seeking because personal source can transfer richer information than impersonal source under most considerations. The distinction between seekers and sources is conceptual in that the same individual can be a seeker or a source at different points in time. This study examines information source preference from the perspective of information seekers because a seeker may use different criteria to evaluate the same information source under different conditions. This study proposes that task importance as a moderator affects the importance of source quality and various costs. This study advances theoretical development on information source preference in two -4- important ways. First, it encompasses cost-benefit factors affecting information source preference. Second, it involves task characteristics to account for how a seeker trades off cost-benefit factors concerning information source preference. The results also suggest perceived content quality should be primarily considered when managers decide to manage organizational information. This thesis is started by reviewing main research on information seeking behavior and setting the scope. Then, focusing on seeker’s information source choice problem, we propose a research model and develop the hypotheses based on integrating cost-benefit framework with elaboration likelihood model. After that, we report an empirical study, which is followed by data analysis and discussion. Finally, we conclude the thesis by presenting the implications of the research results and suggesting directions for future research. -5- 2. Conceptual Background Research on information seeking behavior involved several main fields, including organization communication, information science, decision making, innovation management and consumer research. No coherent definition of information seeking behavior existed in these research areas. However, several similar conceptions regarding information seeking behavior have been discussed in terms of different research interests and contexts. In information science, Wilson (2000) proposed four conceptions concerning information behavior as below: Information behavior is the totality of human behavior in relation to sources and channels of information, including both active and passive information seeking, and information use. Information seeking behavior is the purposive seeking for information as a consequence of a need to satisfy some goal. Information searching behavior is the ‘micro-level’ of behavior employed by the searcher in interacting with information systems of all kinds. Information Behavior Information Seeking Behavior Information Search Behavior Figure2-1: Conceptions relevant to information behavior (Wilson, 1999) -6- In organizational settings, environment scanning is defined as the acquisition and use of information about events and trends in a firm’s external environment, the knowledge of which would assistant management in planning the firms’ future courses of action (e.g., Choo, 2001, 1994; Auster & Choo, 1993; Aguilar, 1967). Besides, feedback seeking behavior regarded as one kind of information seeking behavior under the specific condition is defined as conscious devotion of effort toward determining the correctness and adequacy of behaviors for attaining valued end states (e.g., Ashford & Cummings, 1985). Viewing the newcomer as one specific kind of information seekers in organizations, newcomer information seeking behavior is defined as the extent to which newcomers seek various types of information in terms of requirements (e.g., Morrison, 1993a). The basic problem formulation shared by these studies is how individuals seek information. A main assumption in these studies is that the task leads to information seeking behavior. Though wording of the problem statement varies among the studies, they do conceptually share a common problem field with the central concepts being “information needs” and “use of information sources”. Generally, information seeking is triggered by information needs which are mainly influenced by the task under consideration (e.g., Leckie et al., 1996). Thus, to understand how the task leads to information seeking and consequent information source selection, we firstly review how the task affects information needs which are the cause of information seeking. 2.1 Information Needs Information need is the gap between information seekers’ knowledge about the task under consideration and perceived requirements of the task (e.g., Byström & Järvelin, -7- 1995; Belkin et al., 1982). In organization settings, information need is classified into four categories: task mastery, role clarification, acculturation, and social integration (e.g., Morrison, 1993b). Task mastery refers to one kind of information needs in which individual requires job-related information to perform job tasks; role clarification refers to one kind of information needs in which individual needs information on how others judge him/her and what kind of roles is expected of him/her; acculturation refers to one kind of information needs in which individual needs to understand the organizational norms and values; and social integration refers to one kind of information needs in which individual requires information on how to be involved in the work team. Apparently, different types of information needs are caused by different task requirements, and consequently lead to different information types. 2.1.1 Types of Task Attributes Research on task attributes is relevant to this study in that the interpretation of information needs greatly depends on understanding the task under consideration (e.g., Vakkari, 1999; Ingwersen, 1992; Belkin et al., 1982). A task is an activity to be performed in order to accomplish a goal (e.g., Vakkari, 2003; Hansen, 1999; Hackos & Redish, 1998; Shepherd, 1998), which is defined in two ways: (1) as an abstract construction not including performance (e.g., Vakkari, 2003; Byström, 1999; Shepherd, 1998; McCormick, 1979); (2) as a function of a series of physical and cognitive actions undertaken in pursuit of a particular goal by an actor (e.g., Vakkari, 2003). Two characteristics of the task as determinants of information seeking were suggested in the information seeking behavior literature: task complexity (e.g., Byström, 2002; Vakkari, 1999; Byström & Järvelin, 1995), and task uncertainty (e.g., Ashford & Cummings, 1983; -8- Culnan, 1983). The theoretical difference between task complexity and task uncertainty is beyond the interest of current study. Therefore, in this study, we focus on the distinction between task complexity and task uncertainty at the descriptive level. Firstly, complexity as a central feature of the task has gained great attention in the information seeking behavior literature (e.g., Byström, 2002; Vakkari, 1999; Byström & Järvelin, 1995; Culnan, 1983). Complexity can be understood in many ways (e.g., Campbell, 1988): (1) as primarily psychological; (2) as a person-task interaction; (3) as objective task characteristics. In the information seeking case, task complexity is defined as the average complexity of the most common tasks of each servant, and was reduced to a uni-dimention conception which includes simple vs. difficult or complex task (e.g., Byström & Järvelin, 1995). Simple task refers to which the elements of task are predetermined; while difficult or complex task refers to new and genuine decision task in which the paths of task performance and information needed cannot be determined in advance. For the former, information need may require task-related information for performing the task. For the latter, since the seeker neither knows the paths of task performance nor the task-related information, information need may require all types of information with respect to the task under consideration. Secondly, uncertainty as another important feature of the task has been studied in the information seeking literature (e.g., Zeffane & Gul, 1993; Ashford & Cummings, 1983). In feedback seeking behavior, uncertainty is defined in two ways: (1) as a function of the number of possible responses to stimuli available to an individual as well as their equipotentiality (e.g., Berlyne, 1960); (2) as a function of the sheer number of things that can happen and their equiprobability (e.g., Jones & Gerard, 1967). For the former, -9- information need was to resolve the uncertainty by indicating what most appropriate behavior should be conducted for achieving the desired goal. For the latter, information need was to maintain the uncertainty on certain level by ensuring how the behavior undertaken may be evaluated by others (e.g., Ashford & Cummings, 1983). In addition to task complexity and task uncertainty, task importance may also be a significant determinant of information need. Although no empirical evidence directly suggested that task importance as one of task attributes motivated information seeking, several studies utilized construct that conceptually overlapped task importance provided supportive empirical evidence (e.g., Posavac et al., 2003; Morrison & Vancouver, 2000; Vancouver & Morrison, 1995). For example, need for achievement was measured as a motivator for individuals’ feedback seeking (e.g., Morrison & Vancouver, 2000; Vancouver & Morrison, 1995). Besides, an individual with high accountability had a strong desire for performance information seeking than that with low accountability (e.g., Posavac et al., 2003). Such conceptions are of interest because it is consistent with the definition of personal relevance which consequently leads to task importance. Personal relevant is defined as the extent to which the issue has “intrinsic importance” (e.g., Sherif & Hovland, 1961), or “personal meaning” (Shefit et al., 1973). Prior research on personal relevance in social psychology have linked this construct “ego-involvement” (Rhine & Senverance, 1970; Sherif et al., 1965), “issue involvement” (Kiesler et al., 1969), “personal involvement” (e.g., Sherif et al., 1973). Personal relevance occurs when individuals expect the issue “to have significant consequences for their own lives” (e.g., Apsler & Sears, 1968). Since both need for achievement and accountability are - 10 - related to personal relevance and lead to task importance, task importance can also be regarded as a motivator for information seeking. In summary, three types of task attributes, which are task complexity, task uncertainty and task importance, have been identified in the information seeking behavior literature. All three types of task attributes drive an individual to seek information for simplifying task complexity, resolving task uncertainty, or achieving the task. In addition, task attributes also varies information needs (e.g., Morrison, 1993a; 1993b; Ashford & Cummings, 1985), which results in various types of information sought (e.g., Morrison, 1993b). 2.1.2 Types of Information Sought Previous research in information seeking behavior has presented that information is any stimulus that reaches an individual’s sensory systems (e.g., Miller, 1969), which may be viewed as commodity or things (e.g., recorded knowledge), as knowledge (e.g., personally believed by somebody), as event, and as part of the communication process (e.g., Buckland, 1991). Empirical studies on information seeking have noted several types of information sought (e.g., Byström, 2002; Byström & Järvelin, 1995; Vancouver & Morrison, 1995; Blythe & Royle, 1993; Miller & Jablin, 1991; Ashford & Cummings, 1985). In studying the relationship between task complexity and information source usage, information was classified into three dimensions on a basis of problem solving viewpoint in the cognitive domain: domain information which consists of known facts, concepts, laws, and theories in the domain of the problem; problem information which describes the structure, properties, and requirements of the problem at hand; and problem-solving information - 11 - which covers the methods of problem treatment. All three information categories corresponded different information sources (e.g., Byström, 2002; Byström & Järvelin, 1995). In addition, research on nurses’ information seeking behavior also showed that individuals sought different types of information in terms of task requirements. For example, nurses sought two broad kinds of information for different purposes (e.g, Blythe & Royle, 1993). First, nurses sought patient-specific information which helped care individual patients. Second, nurses also asked institution-specific information or administrative and personnel information for knowing job responsibility, job role and job performance expectations. Moreover, nurses depended on various information sources in terms of the types of information. For example, nurses primarily sought information on patient care from knowledgeable colleague and ward-based sources. However, facing administrative or institutional information, nurses tended to track equipment, reports and people, or locating hospital policies (e.g., Petty & Graves, 1990). Besides, in organization settings, information was classified into four categories (e.g., Miller & Jablin, 1991; Ashford & Cummings, 1985): performance feedback, potential for advancement, appropriateness of social behaviors, adequacy of basic skills and abilities. Such category has been extended into five types (e.g., Morrison & Vancouver, 2000; Morrison, 1993a; Ostroff & Kozlowski, 1992): (1) technical information which is related to performing specific job task or assignment; (2) referent information which explains the role responsibilities and expectations; (3) performance feedback which describes how others perceive and evaluate one’s job performance; (4) social feedback which indicates the acceptability of the non-task behavior or how to behavior within one’s workgroup, and (5) organizational information which describes organizational policies, procedures, - 12 - structure, and objectives. The corresponding relationship between information types and information needs was showed in Figure 2-1. Information Types Technical Information Information Needs Task Mastery Referent Information Role Clarification Performance Information Social Integration Social Information Acculturation Organizational Information Figure 2-1: Information Types and Information Needs (Morrison, 1993a) Generally, a seeker needs to contact different source types to obtain all the information types. All information types have been found important for newcomers to integrate into organizations (e.g., Morrison, 1993a). In addition, it is most likely that several information sources available to a seeker with respect to information needs. For example, to get job task relevant information, a seeker may ask direct supervisor, peers, experience workers, or indirect supervisors. It is not necessary for the seeker to ask every one for information considering cost and time constraints. Consequently, the seeker should make a choice among information sources. - 13 - In summary, prior research on information seeking proposed three types of task attributes and various types of information sought with respect to information seeking. Different information types lead to information source preference of different kinds. Moreover, for the same information content, a seeker may face several different information sources and consequently needs to make decision on information source choice. To examine information source preference, some factors on information source selection have been studied in the information seeking literature. 2.2 Information Source Preference Some previous studies used the term “channel” to refer to information source (e.g., Byström & Järvelin, 1995; Swanson, 1987; Hardy, 1982; Gerstberger & Allen, 1968). For example, both source and channel were used to describe where individuals seek information (e.g., Hardy, 1982). Although the author intent to examine what information sources used by information seekers and why they were used, he provided a list of combining both information channels and information sources when collecting data, such as public or university libraries, other libraries or information collections at your place of work, journals, training sessions, your personal library or files, experts or Forest Service personnel outside of your work unit, and so on. Obviously, both information channel and information source were promiscuous. Essentially, prior empirical studies had differentiated information channel from information source in the information seeking literature (e.g., Chakrabarti et al., 1983). For example, source is defined as the medium in which the knowledge/information is stored and presented, such as books, periodicals, people, film, or magnetic media; whereas channel is viewed as the means by which an information package is moved from one point to another, such as telephone within or out - 14 - organizations, face-to-face discussion, conference or meeting, mail or bookstore. Consistent with previous research, information source differs from information channel, and the former is of interest in the current study. 2.2.1. Categories of Information Sources Information sources that have been studied in the information seeking literature can be classified into two dimensions (e.g., Aguilar, 1967; Choo & Auster, 1993; Culnan, 1983): personal (e.g., colleague, peers, customers, or clients) vs. impersonal (e.g., files, books, final reports, or demos), and internal (e.g., source within work group, people within organizations, or libraries within organizations) vs. external (e.g., source outside organizations). Personal and impersonal sources are literal meaning; internal source refers to those which are available internal (to work organization), while external source refers to those which are available external (to work organization). Choo and Auster (1993) elaborated the categories of information and added some samples (see table 2-2). Table2-2: Typology of Information Sources Internal External Impersonal Internal memos, reports and studies, corporate library, online databases and office automation systems, etc. Newspapers, periodicals, broadcast media, conferences, industry and trade association publications, and communication, etc. Personal Superiors, board members, subordinate managers, and staff, etc. Customers, competitors, government officers, and business associated, etc. Besides, other categories regarding information source have been presented in the information seeking literature, including informal vs. formal (e.g., Pinelli et al., 1993), live vs. documentary, and direct vs. indirect (e.g., Swanson, 1992). Consistently, - 15 - personal source greatly exceeded impersonal source in importance (e.g., Morrison, 1993a; Gerstberger & Allen, 1968). In addition, when asking technical information, the newcomer was found to rely on personal sources rather than written carriers (e.g., Morrison, 1993a). Since personal source can effectively communicate context-related information which is difficult for impersonal sources. In the current study, we focus on personal source rather than impersonal source. 2.2.2 Factors on Information Source Preference A number of factors have been proposed in the information seeking behavior literature influenced information seekers’ source choice (see table 2-3). Table 2-3: Factors on Information Source Selection Factors Reference Perceived Perceived Costs Others Benefit The amount of Ease of use (time, ------information distance, intellective expected effort) Rosenberg, 1967 Gerstberger Allen, 1968 & Technical quality Accessibility Ease of use O’Reilly, 1982 Accessibility, Information specificity quality (accuracy, timeliness, relevance, and account of information obtained) Hardy, 1982 Time-saving Relevance Selectivity Ease of use Source Consulted External personal sources External impersonal sources Internal impersonal sources Degree of External personal Experience sources Internal personal sources Documentary Task sources complexity Personal sources Task Others uncertainty Individual (motivation, job tenure, and education) ------------ Documentary sources Personal sources - 16 - Promptness Charkrabarti et Utility of stored Ease of use al., 1983 information Availability Skill to use -------- Documentary sources Personal sources Culnan, 1983 ----------- Perceived accessibility Perceived task Personal sources complexity Documentary sources Culnan, 1985 ----------- Physical dimension Interface dimension Tiamiyu, 1992 ---------- --------- Choo, 1993 Perceive source quality (relevance, reliability, comprehensiven ess, timeliness) Perceived source accessibility (time and effort, physical proximity, financial, ease of use) Choo, 1994 Perceived information quality ---- Computer-based Experience sources with source Library Contextual Personal sources factors Human sources Work Documentary complexity sources Decisionmaker discretion Activity duration Scope of External personal scanning sources Amount of External scanning published sources Internal personal sources Internal printed sources ---External personal sources External impersonal sources Internal personal sources Internal impersonal sources Internal channels External channels Need for Personal sources achievement Documentary sources Baldwin Rice, 1997 Morrison Vancouver, 2000 & Individual characteristics Institutional resources & Information Information importance difficulty Source expertise Source accessibility Hertzum & Appropriate Cost/time Pejtersen, 2000 Relevance Accessibility Confidential Availability Intellectual effort Memory Internal personal sources External personal sources Internal impersonal - 17 - Anderson et al., Accuracy 2001 Comprehensive Relevant Ease of use Technical quality Inexpensive Obtainable nearby Easy to obtain Byström, 2002 ---------- ----------- Task complexity Task uncertainty Experience with source Source importance Task complexity sources Internal personal sources Internal impersonal sources People Documentary Hertzum, 2002 Accessibility Ease of use Cost to use Background People source information Document about the sources source Borgatti Cross, 2003 Ability to access (proximity, cognitive) Cost (interpersonal risks and obligations) Knowing (awareness) Tenure, gender, hierarchy Technical quality Up-to-dateness Representativenss Appropriateness to task Appropriate project experience Appropriate organizational unit & Value (expertise) Hirsh & Least time to track down Dinkelacker, Most convenient at time/place of my choosing 2003 Most current-I need the most up-to-date information possible Most authoritative---gives the most reliable Complete information, most familiar—tried and true has worked for me in the past Most reliably available---no waits or hassles Work roles Accessibility Kwasitsu, 2003 Technical quality Technical Availability Relevance jargon Ease of use Currency Experience Cost of use Reliability Personal mastery Yitzhaki & -----------Hammershlag, 2004 Physical accessibility --------- People Computer-based sources Personal sources Internal personal sources Internal impersonal sources External impersonal sources External personal sources Documentary sources Personal sources - 18 - What factors influence a seeker’s information source choice? A cost-benefit diagram is widely adopted by both organizational and information science researchers. However, there is disagreement on the relative importance of the two components. Basically, two streams of research have been identified in the information seeking literature: source quality as more important (e.g., Morrison & Vancouver, 2000; Vancouver & Morrison, 1995; Ashford, 1986), and the least effort model (e.g., e.g., Yitzhaki & Hammershlag, 2004; Anderson et al., 2001; Chakrabarti et al., 1983; Culnan, 1983; Gerstberger & Allen 1968). Perceived quality as a dominant factor influencing information source selection is intuitive. Firstly, it is generally agreed upon that an important objective of information seeking is to resolve uncertainty (e.g., Morrison, 2002; Morrison & Vancouver, 2000; Vancouver & Morrison, 1995; Miller & Jablin, 1991; Comer, 1991; Burke & Bolf, 1986; Ashford, 1986), and to improve task competence (e.g., Tan & Zhao, 2003; VandeWalle et al., 2000; Ashford, 1986; Morrison, 1993a; 1993b); therefore, a source is preferred when it offers quality information. Secondly, the basic assumption for information selection under cost-benefit diagram is that the marginal benefit exceeds the marginal cost of information seeking for the next source (Orr, 1970; Stigler, 1961). Therefore, the cost component will be meaningless if quality is not a concern in information seeking. Both theoretical and empirical studies have supported that perceived quality is important for information seeking. For example, newcomers faced high levels of uncertainty regarding the task role and a sense of acceptance in the groups and organizations, which resulted in information seeking (e.g., Miller & Jablin, 1991; Ashford, 1986; Ashford & Cummings, 1983). Several information sources were available for a newcomer’s - 19 - information needs, such as direct supervisor, co-workers, experienced colleagues, and other sources outside organizations. The newcomer tended to ask technical information from experienced colleagues rather than from direct supervisor because experienced colleagues were familiar with newcomer’s task and technical information provided by them was believed higher quality than that provided by supervisor (e.g., Burke & Bolf, 1986; Comer, 1991). Consistent with prior research, recent study on information seeking also found individuals in organizations tended to rely more on sources with high quality of information for resolve both task and role uncertainty (e.g., Morrison & Vancouver, 2000). In organizations, since high quality of information enabled a seeker to clarify the job expectations, evaluate the adequacy and appropriateness of the work behavior, and improve task competence (e.g., VandeWalle et al., 2000; Morrison, 1993b), the seeker likely focuses on information quality of the source when seeking information. In contrast, the least effort principle naturally emerges from empirical studies rather than as a result of theoretical reasoning. It came as a surprise even to the first a few advocators of this argument (e.g., Gerstberger & Allen, 1968). Why do seekers consider accessibility before source quality? Firstly, the insignificance of source quality might be due to lack of variance in quality among available information sources (e.g., Swanson, 1987; Allen, 1977; Orr, 1970; Rosenberg, 1967). Under the assumption of benefit-cost framework, when the perceived quality of information is equal, the source with the lowest cost will maximize the benefit-cost ratio as a result of information source preference. Secondly, source accessibility as a dominant factor for information source selection rather than perceived quality might be due to the time and resources limitation (e.g., Hertzum & Pejtersen, 2000; Pinelli et al., 1991; Chakrabarti et al., 1983; O’Reilly, - 20 - 1982). In organization settings, individuals may be restricted by either time or awareness of information sources (e.g., O’Reilly, 1982). Under time pressure, individuals tended to rely more on personal sources (e.g., colleagues) due to source accessibility (e.g., Anderson et al., 2001). When individuals do not know where to get high quality of information, it is reasonable that more accessible information source would be selected than less accessible one (e.g., Hertzum & Pejtersen, 2000; O’Reilly, 1982). Though these studies that identify the importance of source accessibility on information source preference, it should be recognized that they fail to clearly measure the dimensions of accessibility on personal source. Furthermore, these investigations also neglect to inquire about the different measures of source accessibility from those of perceived quality. Other research on information seeking found source accessibility was a multidimension conception (e.g., Culnan, 1984; 1984). Source accessibility is defined as both the social and economic costs associated with acquiring information (e.g., Culnan, 1984), which included three dimensions: physical accessibility, system accessibility and information accessibility. Physical accessibility or called terminal accessibility refers to which individuals physically access to terminal, enter a library or locate the target person and gain his/her attention; system accessibility or called interface accessibility refers to any translation of a request into “non-natural” language. For library, this means locating the call number of the item; for computer-based information, this means use of a command language to formulate a query; information accessibility refers to the ability to physically retrieve the information impendent of any subsequent judgment as to the item’s relevance. For computer-based sources, three dimensions of source accessibility had clear differentiation. However, for personal source, these dimensions might not - 21 - enough to measure source accessibility. For example, system accessibility was found to insignificantly affect personal source selection (e.g., Culnan, 1985). In addition, social risk was proposed to be important as a cost factor for information seeking from personal sources (e.g., Tan & Zhao, 2003; Ashford, 1986). On the other hand, some previous studies on the dimensions of source accessibility in the information seeking literature involved the measures of perceived quality. For example, recent research on source accessibility (e.g., Fidel & Green, 2004) found that there were many faces of accessibility, including “sources in know”, “has a lot of different types of information in one place”, “can give the right level of detail”, “saves time”, “has the right format”, “sources with which I feel comfortable”, “is physically close”, “can be searched with keywords or codes”, “is interactive”, “is available”, “is not busy and is accessible”. Obviously, in those faces of source accessibility, some measures of perceived information quality were involved such as “has a lot of different types of information in one place” (quantity), “can give the right level of detail” (scope), and “has the right format” (format). According to the vague measures of source accessibility, we believed more research should be conducted to examine source accessibility. Although both perceived quality and source accessibility were reported very important for information source preference in two research streams, they could not ensure the success of information source selection without considering other factors. Fortunately, more factors which influenced information source selection have been identified in the information seeking literature: individual characteristics (e.g., Borgatti & Cross, 2003; Kwasitsu, 2003; Anderson, et al., 2001; Ashford, 1986; Culnan, 1985; O’Reilly, 1982; Gerstberger & Allen, 1968), institutional resources (e.g., Baldwin & - 22 - Rice, 1997), and task attributes (e.g., Byström and Järvelin, 1995; Tiamiyu, 1992; Hart & Rice, 1991; Culnan, 1983; O’Reilly, 1982; Tushman, 1978). Research on individual characteristics with respect to information source selection mainly included job tenure (e.g., Ashford, 1986; O’Reilly, 1982), organizational tenure (e.g., Ashford, 1986), work role (e.g., Yitzhaki & Hammershlag, 2004; Kwasitsu, 2003; Leckie et al., 1996), education (e.g., Baldwin & Rice, 1997), and experience with source (e.g., Borgatti & Cross, 2003; Kwasitsu, 2003; Anderson, et al., 2001; Kuhlthau, 1999; Culnan, 1985; Gerstberger & Allen, 1968). For example, individuals with long job and organizational tenure were less dependent on information sources than those with short job and organizational tenure because the former likely endured higher social cost if they asked others for job-related information than the latter (e.g., Ashford, 1986). Secondly, individuals with different work roles were also reported to vary in information source selection (e.g., Yitzhaki & Hammershlag, 2004; Leckie, et al., 1996). According to different work roles, engineers in design and development tended to rely more on internal sources because they worked for designing, testing, manufacturing, and constructing new products or services which were generally constrained by time and confidentiality. While engineers with consulting role tended to rely on external sources such as market information about vendors and customers because they worked as contacting with clients, presentations, and supervision of other engineers and technical personnel (e.g., Leckie, et al., 1996). According to differences between academy and industry, scientists heavily used academy journals and conference proceedings as well as information database since they sought information for creating and publishing new and original knowledge. However, engineers relied more on internal personal sources (e.g., colleagues and - 23 - supervisors) because of more accessible and quick feedback (Yitzhaki & Hammershlag, 2004). Thirdly, engineers with high level of education were found not to be personal mastery and tended to seek information from libraries rather than own personal memories (e.g., Kwasitsu, 2003). Oppositely, educated individuals were found to infrequently depend on other sources because they possessed information required (e.g., O’Reilly, 1982). More interestingly, education was also found not relevant to information seeking by security analyst (e.g., Baldwin & Rice, 1997). Finally, prior experience with the source was proposed to be related to information source selection (e.g., Borgatti & Cross, 2003; Kwasitsu, 2003; Anderson et al., 2001; Kuhlthau, 1999; Baldwin & Rice, 1997; Culnan, 1985; Gerstberger & Allen, 1968). For example, a seeker was reported to likely choose the information source in which the seeker successfully obtained information needed before rather than look for new information sources because of self-reinforcing (e.g., Kuhlthau, 1999; Baldwin & Rice, 1997; Allen, 1977). Yet, comparatively studying the impacts of both individuals’ characteristics and institutional resources on information seeking noted that individuals characteristics did not affect information seeking by security analysts except for years worked, but institutional resources, such as staff size, budget (time/ budget limit resources/conferences), type of firm, and location of firm, were found to affect security analysts’ information source selection (e.g., Baldwin & Rice, 1997). Besides, as mentioned above (section 2.2.1), three types of task attributes have been identified in information seeking behavior: task complexity, task uncertainty and task importance. Task complexity and task uncertainty were proposed to be important features in determining information seeking (e.g., Byström & Järvelin, 1995; Tiamiyu, 1992; Hart - 24 - & Rice, 1991; Culnan, 1983; O’Reilly, 1982; Tushman, 1978) inasmuch as they steered information needs and consequently information seeking. For example, the number of information sources increased as task complexity increased (e.g., Anderson et al., 2001; Byström & Järvelin, 1995). On the other hand, individuals frequently seek information as the task uncertainty increased (e.g., Morrison, 1993a; 1993b; Ashford & Cummings, 1985). For example, two kinds of task uncertainty were proposed in research on feedback seeking (e.g., Ashford & Cummings, 1985): role ambiguity and contingency uncertainty. Role ambiguity refers to the uncertainty about the performance of the individual job role, the nature of job responsibilities and the expectations of others for behavior in that job, such as lack of information regarding supervisory evaluations of the work, opportunities for advancement, and the expectations of role senders (e.g., Ashford & Cummings, 1985). Contingency uncertainty is defined as the individual’s experienced uncertainty about the links between evaluations of current performance and the achievement of second other outcomes. For example, individuals were lack of information regarding organizational context, and promotions or career success. To resolve such uncertainty, a newcomer was found to frequently seek information. Moreover, individuals might vary information source selection in terms of task requirements. For example, senior managers were found to rely on multiple sources: personal sources were important for getting information on market and environment outside organizations, whereas documentary sources were important for getting information on technological and regulatory matters (e.g., Auster & Choo, 1994). However, both task complexity and task uncertainty were also found not to be relevant to information source selection by decision makers (e.g., O’Reilly, 1982). More interestingly, few empirical studies in information seeking behavior concerned task - 25 - importance on information source selection (1995; Morrison & Vancouver, 2000). For example, high achievers were found to rely on the sources with quality of information than low ones (Morrison & Vancouver, 2000). These studies identified how task attributes affected information source preference by information seekers, but miss considering how task attributes affects individuals’ trade-off between perceived quality and source accessibility. For example, Morrison and Vancouver (2000) did not examine whether need for achievement affected the importance of source accessibility on information source preference. This omission seems surprising because information source preference appears to be based mostly on balancing between perceived quality and source accessibility. Given differences between perceived quality and source accessibility in terms of importance, a seeker would seem likely to vary information source preference by trade-off between perceived quality and source accessibility. In summary, the purpose of this thesis is to determine the factors influencing information source preference, and investigate if task importance affects the importance of those factors on information source preference. - 26 - 3. Research Model and Hypotheses Development The dependent variable in this study is personal source preference from which information seekers will ask task-relevant information. The factors influencing information source preference are generated by cost-benefit framework (Orr, 1970; Stigler, 1961), and interaction effect is developed based on Elaboration Likelihood Model (Petty & Cacioppo, 1986; 1994). 3.1 Cost-benefit Framework The cost-benefit framework was raised out in economics (e.g., Urban et al., 1993; Bettman, 1979; Stigler, 1961), which proposed that individuals search for information until the marginal cost of obtaining a unit of information is equal to the marginal benefit of possessing a unit of information. Thus, information search will decrease as the costs of searching increase, and will increase as the benefits of searching increase. In the context of information source preference, information seekers choose the target person by balancing perceived benefits and perceived costs. Perceived benefit is defined as outcome that increases one’s utility or provides value by facilitating achievement of higher level goals (e.g., adapted from Schmidt & Spreng, 1996; Olashavsky & Wymer, 1995; Gutman, 1982). In this case, we focus on one benefit factor: perceived content quality. Perceived content quality refers to the extent to which information seekers believe the content received fitness for use. In addition, perceived cost is defined as the individual’s subjective assessment of the monetary expenditure, time sacrifice, physical effort, and psychological sacrifice that he or she expends searching for information (Schmidt & Spreng, 1996). Unlike perceived benefit which pertains to information content alone, cost may pertain to both information content and the source which carries - 27 - the content. The cost pertaining to content can be summarized into source understandability (e.g., Xu & Chen, 2005; Swanson, 1982). The cost pertaining to the access to the source per se is reflected in physical proximity (Fidel & Green, 2004; McCreadie & Rice 1999; Chakrabarti et al., 1983; Gerstberger & Allen 1968), ease of use (Kankanhalli et al., 2005; Culnan, 1984; Chakrabarti et al., 1983; Hardy, 1982) and social risk (VandeWalle et al., 2000; Ashford, 1986). Ease of use applies only to impersonal sources such as document and digital repositories. For documents, ease of use can be affected by the use of right format (Fidel & Green, 2004). For digital repositories, it is affected by the ease of retrieval (Hertzum & Pejtersen, 2000) or the use of retrieval language (Culnan, 1985). Therefore, in the personal source selection case, three cost factors are involved: physical proximity refers to the linear distance between the source and the seeker; source understandability refers to the ease with which a seeker can decode and understand the information (Xu & Chen, 2005); social risk refers to the embarrassment, loss of face, and revealing of incompetence (Ashford, 1986). 3.2 Elaboration Likelihood Model The term elaboration likelihood refers to the likelihood one engages in issue-relevant thinking with the aim of determining the merits of the arguments for a position rather than the total amount of thinking per se in which a person engages (e.g., Petty & Cacioppo, 1986) The elaboration likelihood model (ELM) posits that information processing is motivated by personal relevance/involvement. It asserts that when people are motivated to engage in evaluating issue-relevant messages, it is more likely that they scrutinize the messages content and base their judgment on the merits of the content rather than that of - 28 - the source. Such type of processing is called the central route of processing. In contrast, if people are unwilling to process information items, less cognitive capacity will be devoted; resulting in a judgment based more on peripheral cues. Hence, which cue individuals choose to process message partly depends on the motivation. One of major motivation variables is personal relevance or called involvement. Involvement is defined as the psychological state triggered by two key aspects of an issue—its importance or significance and its personal relevance (Barki & Hartwick, 1989, 1994; Petty & Ciacoppo, 1979). This conception was extended by Johnson and Eagly (1989), and recently applied in information seeking behavior. In the information seeking case, three dimensions of involvement have been identified: value-relevant, outcome-relevant and impression-relevant involvement (e.g., Cho & Boster, 2005). Value-relevant involvement is defined as the psychological state that is created by the activation of attitudes that are linked to important values (Johnson & Eagly, 1989); outcome-relevant involvement refers to the consequences of issue under consideration is of personal importance (Johnson & Eagly, 1989); impression-relevant involvement is defined as the individual’s concern with the consequences of his response or with the instrumental meaning of his opinion (e.g., Zimbardo, 1960; Johnson & Eagly, 1989). The former two dimensions are relevant to the issue itself, whereas the latter one is relevant to the consequences of communication rather than the issue. Hence, we developed the conception of task importance based on value-relevant and outcome-relevant involvement. Task importance is defined as the degree of which information seekers perceived significance or personal relevance of the task under consideration based on inherent needs, values, and interests. - 29 - According to ELM, individuals with high involvement tend to carefully and thoughtfully consider the true merits of the information, while individuals with low involvement tend to rely on source cues rather than message cues for saving cognitive efforts. Although the elaboration likelihood model is widely used to explain people’s cognitive processing of persuasion message, it has been suggested to be applicable to information seeking (e.g., Cho & Boster, 2005; Posavac & Herzenstein, 2003). In the information seeking case, value-relevant and outcome-relevant involvement were found to be related to information seeking (e.g., Cho & Boster, 2005). In this study, regarding information source preference, we believed that task importance caused by both value-relevant and outcomerelevant involvement also affected information seeking and even the weight of perceived costs and perceived benefits. 3.3 Research Model The research model explaining information source preference with corresponding factors is showed below (see Figure 3-1). Previous research has emphasized the importance of perceived content quality (e.g., Morrison & Vancouver, 2000; Vancouver & Morrison, 1995; Ashford, 1986) and physical proximity (e.g., Field & Green, 2004) on determining information source preference. Source understandability and social risk are also two determinants of perceived costs of information source preference. Prior studies have also indirectly emphasized task importance on moderating individuals’ information source selection (e.g., Morrison & Vancounver, 2000). Therefore, perceived content quality, physical proximity, source understandability, and social risk are hypothesized to impact individuals’ personal source preference, which are moderated by task importance. - 30 - Perceived Content Quality Information Source Preference Physical Proximity Source Understandability Social Risk Task Importance Figure 3-1: Research Model of Personal Source Preference 3.3.1 Perceived Content Quality Perceived content quality refers to the extent to which individual believes the content of information itself sought is fit for use. Prior research suggested perceived content quality is a significant determinant of information source selection by the seekers, and several main dimensions involved (e.g., Xu & Chen, 2005; Moenaert & Sounder, 1996; Vancouver and Morrision, 1995; O’Reilly, 1982). O’Reilly (1982) proposed several main aspects of information quality regardless of personal or impersonal sources: accuracy, relevance, specificity, reliability, and timeliness. Expertise was represented as the determinant of information quality regarding asking information from personal sources (Vancounver & Morrison, 1995). Moenaert and Sounder (1996) identified four main information dimensions influencing information utility during communication across departments: relevance, novelty, credibility, and comprehensibility. Recent research by Xu and Chen (2005) suggested topicality, novelty, understandability, scope, and reliability as the main dimensions influencing document quality. In the context of - 31 - information source preference, perceived content quality contains relevance, scope, and novelty. H1a: Perceived content quality is positively related to seekers’ information source preference. The positive relationship between perceived content quality and information source preference may be strengthened by task importance. When the consequence of the task is important or personal relevant to the seekers, the seekers tend to require information with high quality since such information may get the task under consideration performed effectively and efficiently. Prior research suggested several important consequences for the seekers: momentary rewards, social rewards (e.g., Ashford, 1986), or even intrinsic rewards (e.g., Morrsion & Vancounver, 2000). For example, need for achievement as intrinsic reward moderated information source preference with task expertise (Morrsion & Vancouver, 2000). H1b: When seekers perceive higher task importance, they place more weight on perceived content quality than when they perceive lower task importance. 3.3.2 Physical Proximity Physical proximity is defined as the degree of the distance between the seeker and the source. Previous research suggested information source selection involved physical proximity. Information seekers are wanted to track down the target person who possesses the information required. The more accessible source tends to be located near the seekers - 32 - (e.g., Field & Green, 2004). For example, engineers in charge of production and services prefer asking their colleagues to people outside work group (e.g., King et al., 1994). H2a: Physical proximity is positively related to seekers’ information source preference. When the task under consideration is perceived important, information seekers pay less attention to physical proximity. According to elaboration likelihood model, when the issue is perceived important or personal relevance, individuals are willing to devote more physical efforts to processing issue-relevant message. In the case of information source preference, information seekers do not care about taking time or traveling longer distance to approach the target person who can provide the task relevant information which can be used to handle the task whose consequences are perceived important. H2b: When seekers perceive higher task importance, they place less weight on physical proximity than when they perceive lower task importance. 3.3.3 Source understandability Source understandability is viewed as a cognitive cost (Xu & Chen, 2005; Swanson, 1982) affecting information source preference due to the seekers need take cognitive efforts to process the information sought. To understand new technical information, information seekers use prior knowledge to convert the new content into knowledge matching their cognitive structure. If information obtained by the seekers meets the seekers’ knowledge domain or is close to their cognitive structure, then the seekers would take few cognitive efforts to apprehend what they obtained from the source. Consequently, the seekers prefer the source which costs fewer efforts. - 33 - H3a: Source understandability is positively related to seekers’ information source preference. The elaboration likelihood model suggests that if the consequences of the issue are perceived important or personal relevant, individuals will be motivated to process issue information with more cognitive efforts. The relationship between source understandability and information source preference may be contingent on the condition of task importance. If the consequences of the task under consideration are perceived important, seekers are willing to exert more cognitive efforts to understand the information sought even if it is not explained clearly, or goes beyond the seekers’ knowledge domain. H3b: When seekers perceive higher task importance, they place less weight on source understandability than when they perceive lower task importance. 3.3.4 Social risk Social risk refers to the embarrassment, loss of face, and revealing of incompetence (e.g., Ashford, 1986) when one asks a personal source for information. For example, oldtimers in an organization are supposed to know the organizational rules and not to ask for them; while for new-comers it is perfectly legitimate. Another example is when one performs unsatisfactorily in job; seeking performance feedback can entail embarrassment. H4a: The social risk entailed in approaching a source is negatively related to seekers’ information source preference. - 34 - When the consequences of the task under consideration are perceived important or personal relevant, information seekers need to care whether they can achieve the task. Under such circumstance, information seekers would like to try their best to seek information required. In a way, information seekers with learning orientation tend to view information seeking as an approach to improve performance rather than a signal of incompetence (e.g., WandeWalle & Cummings, 1997). H4b: When seekers perceive higher task importance, they place less weight on social risk than when they perceive lower task importance. - 35 - 4. Research Methodology Consistent with prior research design in the information seeking behavior literature and due to the generalizability of the results (Dooley, 2001), the survey method was used in this study. Research instruments were first of all generated and developed by carefully searching the domain of the constructs. Two experts in IS research area reviewed the instruments to assess the face validity. Secondly, a pilot study was conducted for measuring the reliability and validity of instruments. Main survey was then performed to collect data after ensuring the quality of the survey instruments. Finally, statistical techniques are used to test the research model: CFA for measurement model and MMR for structure model as well as interaction effect. 4.1 Item Generation and Content Validity The purpose of item generation was to identify the items to fit the construct definitions. Our predictor measures (dependent variables) were perceived content quality, physical proximity, source understandability, and social risk, and criterion variable (independent variable) was personal source preference. Task importance as a moderated factor (interaction effect) was measured in this study. We used multi-item scales because single item failed in representation of the construct. Independent variable Information source preference refers to the extent to which information seekers make a choice among interpersonal sources for task-related information. In the information source preference case, we used self-developed items. Past literature (e.g., O’Reilly, 1982; Vancouver & Morrison, 2000) in organizational studies almost exclusively measured frequency of information seeking from a specific source. We - 36 - consider frequency a measure of amount of seeking rather than choice. One may infrequently use a source because the problem situation is infrequent. Therefore, amount of seeking is less appropriate. Source preference is measure with (1) how strong a seeker considers a source as the best source and (2) how much the seeker prefer to ask the source for knowledge among all sources. In order to generate sufficient variance among respondents for the presence of moderators, the dependent variable was measured using 100 point Likert scale. Dependent variables Perceived content quality refers to the extent to the content of information sought is fit for use. Six-item perceived content quality, borrowed from previous research, was assessed in the study. These measures operationally define the general domain of jobrelated information quality, without regard to their prevalence across professionals. Respondents were asked to state the extent to which they agree (or disagree) the statements describing their personal sources. Seven-point Likert scales were used for this purpose. Physical proximity refers to extent to how far or how close information seekers are located from information sources from which information seekers can obtain the information needed. It also can be called physical distance, which has been operationalized in the literature primarily in terms of geographic regions (Almeida, 1996) and geographic proximity (Galbraith, 1990; Lester and McCabe, 1993). In this case, respondents were provided with a number of statements and were asked to indicate (on a seven-point scale) to what extent they agree that each statement represents personal source selected with considering the specific job-related task. The statements used were - 37 - of the following types: “His/her office is located close to mine”; “I do not have to travel a long distance to his/her office”; and “it is not difficult to approach him/her in person”. Source understandability is defined as the ease with which a seeker can decode and understand the information. In this study, four questionnaire items were used to measure this construct. Amongst two of four were adopted from prior study (Xu & Chen, 2005), and another two items were self-developed. Each item was weighed against a seven-point Likert scale denoting the extent to which respondents agree or disagree that the statements describe their job-related task. The scales ranged from strongly disagree (scoring 1) to strongly agree (scoring 7). Social risk refers to the embarrassment, loss of face, and revealing of incompetence (Ashford, 1986). In this study, we borrowed the work by Ashford (1986) and used four items with seven-point Likert scale. Respondents were asked to state the extent to which they agree (or disagree) the statements describing their personal sources. Moderated variable Task importance refers to the extent to which individuals perceive significance or personal relevance of the task under consideration based on inherent needs, values, and interests. The items were presented in a seven-point Likert-type format with the response scale ranging from strongly disagree to strongly agree. The items were scored so that a higher number indicated higher task importance. Control variables Following research on feedback seeking behavior (e.g., O’Reilly, 1982; Ashford, 1986), we measured control variables using education, gender, age, job tenure, and seekers’ prior knowledge. Seekers’ prior knowledge refers to extent to which a person - 38 - has an organized structure of knowledge (schema) concerning an issue. The items were presented in a seven-point Likert-type format with the response scale ranging from strongly disagree to strongly agree. In summarize, perceived content quality (PCQ) was an eight-item scale measuring the extent to the fitness for use of job-related information sought. Physical proximity (PP) was a three-item scale measuring the extent to the physical distance between personal source and information seekers. Source understandability (SU) was a four-item scale measuring the extent to which individuals would like to devote attention to involving the information sought. Social risk (SR) was a four-item scale measuring the extent to which individuals can take the psychological pressure when asking others for knowledge. Task importance (TI) was three-item scale measuring the extent to which knowledge workers psychologically evaluate the concerned task. Seekers’ prior knowledge (PK) was fouritem scale measuring the organized knowledge structure related to the task under consideration. Information source preference (ISP) was four-scale measuring the extent to which individuals make the choice of personal sources for job-related information. Respondents were knowledge workers across professionals. - 39 - Table 4-1: Item Generalization Construct Items Item wording PCQ1 PCQ2 PCQ3 Perceived quality content PCQ4 PCQ5 PCQ6 PCQ7 PCQ8 PP1 Physical proximity PP2 PP3 SU1 Source understandability SU2 SU3 SU4 SR1 SR2 Social risk SR3 SR4 References Adapted from He/She has knowledge that is (McKinney et al., potentially applicable to the task. 2002) Adapted from He/She has knowledge that is (McKinney et al., relevant to the task. 2002) Adapted from He/She has broad knowledge (McKinney et al., related to the task. 2002) He/She knows the task well. Self-generated Developed based on He/She is an expert in the task (Morrison & Vancounver, 2000) He/She has unique knowledge Adapted from (Xu & which can be sued to conduct the Chen, 2005) task. He/She has new knowledge Developed based on before other people. (Xu & Chen, 2005) He/She has knowledge that is Adapted from (Xu & new to me. Chen, 2005) His/her office is located close to Developed based on mine. (Fidel & Green, 2004) I do not have to travel a long Developed based on distance to his/her place. (Fidel & Green, 2004) It is not difficult to approach him Self-generated in person. He/She is able to explain the Self-generated issue clearly for me. The knowledge from him/her is Adapted from (Xu & easy to understand. Chen,2005) I can easily follow that he/she Adapted from (Xu & suggests. Chen,2005) The knowledge from him/her is Self-generated clear in meaning. I would be nervous to ask Adapted from him/her for task knowledge. (Ashford, 1986) It is embarrassing to ask him/her Adapted from for ask knowledge. (Ashford, 1986) He/she might think I am Adapted from incompetent if I ask him/her. (Ashford, 1986) I think he/she would think worse Adapted from of me if I ask him/her. (Ashford, 1986) - 40 - TI1 Task Importance TI2 TI3 PK1 Prior Knowledge PK2 PK3 PK4 ISP1 Information source preference ISP2 ISP3 ISP4 The task is an important part of my duty. The task is important to my performance. Developed based on (Zaichkowsky, 1985) Developed based on (Zaichkowsky, 1985) Developed based on The task means a lot to me. (Zaichkowsky, 1985) I consider myself an expert in Adapted from (Xu & doing this task. Chen, 2005) I can tell people a lot of how to Adapted from (Xu & do this task. Chen, 2005) Adapted from (Xu & I know this task very well. Chen, 2005) I logically analyze this task. Self-generated If I come across problem in the task, I think he/she is the best Self-generated personal to approach for problem solving knowledge. Among all the people available to me, I prefer to ask him/her for Self-generated task knowledge. Without him/her, it would be more difficult for me to obtain Self-generated needed knowledge in case of problems Overall, he/she is a very useful Self-generated source of task knowledge. - 41 - 4.2 Scale Evaluation Given that the literature suggested a considerable overlap in the construct domain for perceived content quality and perceived content understandability, a pilot study was conducted and exploration factor analysis was applied to empirically confirm the independent dimensionality of the scales. Sample in the pilot study was chosen within a school of a top-tier university, including 44 graduate students with work experience in a class and 21 technical staff working in the school. Factor analysis with Varmax rotation was conducted to assess construct validity. Seven factors with Eigenvalues greater than one were obtained (see table 4-2). A twostage rule was used to categorize items to factors (Nunnally, 1978). On the one hand, to ensure a given item represented the underlying factor, we checked whether each item loading fall in the acceptable range. Commrey (1973) indicated that loadings in excess of .45 could be considered fair, greater than .55 good, .63 very good, and .71 excellent. Most items were loaded on the target factors in the excellent range except of PCQ 7 and PCQ 8 loaded heavily on another latent factor and were omitted. On the other hand, to avoid problems with cross-loadings, the difference between weights for any given item was checked less than .10 across factors. In summary, these results indicated that the various scales achieved the reliability. - 42 - Table 4-2: Rotated Component Matrix (a) Rotated Component Matrix Component 2 3 4 5 .356 -.008 .056 .117 .313 -.032 .043 .196 .233 -.139 .064 .140 .242 -.014 -.035 .062 .053 -.012 .078 -.004 .070 -.187 -.005 .171 .241 -.012 -.214 .057 .093 -.158 .097 .075 .450 -.184 -.114 .079 -.008 .045 .032 .709 -.113 .033 .189 .825 -.110 .072 .158 .849 -.196 .072 .091 .837 -.001 .061 .040 .881 -.066 .077 .000 .933 -.089 .041 .015 .913 -.183 -.025 -.132 .860 .092 -.062 .184 .864 .074 -.070 .170 .920 .181 .059 .097 .900 .264 -.127 .011 -.052 .396 -.183 -.049 -.007 -.086 .361 .083 -.126 .160 .013 -.111 .167 -.206 -.001 .082 .876 .017 .011 .137 .890 .102 .092 .103 .902 .316 .076 .122 .757 4.07 3.43 2.25 1.72 1 6 7 PCQ1 .236 .085 .748 PCQ2 .168 .085 .849 PCQ3 .189 -.014 .837 PCQ4 .138 .087 .880 PCQ5 .144 -.043 .896 PCQ6 .189 .058 .722 PP1 -.031 .098 .847 PP2 .165 -.032 .865 PP3 -.003 .006 .734 SU1 .310 .236 .159 SU2 .253 .119 .296 SU3 .271 .123 .138 SU4 .307 .075 .202 SR1 -.087 .032 -.138 SR2 -.061 .047 -.005 SR3 -.097 .011 -.047 SR4 -.057 -.060 -.131 TI1 .194 .082 .147 TI2 .158 -.058 .133 TI3 .115 -.043 -.068 ISP1 .420 -.109 .747 ISP2 .298 .054 .742 ISP3 .137 .072 .718 ISP4 .417 .062 .776 PK1 -.091 .034 -.071 PK2 .002 -.014 .046 PK3 .119 .016 -.065 PK4 .120 -.075 -.099 Eigenvalue 9.28 1.23 1.15 Variance 33.13% 14.55% 12.25% 8.04% 6.13% 4.38% 4.11% extracted Cumulative 33.13% 47.68% 59.93% 67.97% 74.10% 78.48% 82.59% varaince Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. a Rotation converged in 7 iterations. - 43 - 4.3 Survey Administration The sample for main survey was comprised of knowledge workers across professionals, such as engineers, nurses, administrative officers, educators, and librarians. Although information seeking behavior of these professionals has been studied separately in the information seeking literature, there is no empirical study involving all the professionals in one research. In fact, Leckie et al., (1996) proposed a general model which covers most of the professionals, but no empirical study approves the model in the literature. A potential pool of 200 participants was drawn covering several professionals (see table 4-3): engineers, educators, nurses, and librarians. We chose these field segments for their acknowledged knowledge required in terms of work requirements. A senior manager in each organization was initially contacted by sending a cover letter and questionnaire measuring knowledge seeking behavior (Appendix A) for getting the approval of collecting data from his/her department. Then, the questionnaires with cover letter were distributed either by the researcher or by the senior manager. Four sections involved in the survey questionnaires. In the first section, respondents were first of all required to list four job-related tasks classified into two groups: frequent task and occasional task. These four tasks should be those to which respondents are lack of knowledge related. Then, for building personal source pool, respondents were required to write down at least four persons (six was maximize) whom respondents need to contact at least once in a while for job tasks. In the second section, we used the month of respondents’ birth to randomly choose one person among the personal source pool built in the first section. In the third section, referring to one appointed task listed in the first section, respondents were required to answer the statements regarding the personal - 44 - source selected in the second section. The background information was arranged in the final section of the questionnaire. All respondents were motivated by offering monetary rewards, and one of organizations was motivated by offering the research result. A final total of 161 responses (85 percent) were obtained which should be sufficient to obtain an accurate solution in factor analysis as long as item interrelations are reasonably strong (Guadagnoli & Velicer, 1988). Table 4-3: Characteristics of Participating Respondents Demographic Age Gender Education Work role Job tenure Industry tenure Category 20-30 31-40 41-50 51-60 Male Female Middle/high school Diploma Undergraduate Graduate PHD Others Managers Administrative officers Engineers Teachers Nurses Students Librarians Others =10 =20 Frequency (n=161) 64 55 28 6 40 120 35 19 37 64 1 5 21 51 17 4 37 1 19 11 42 73 33 11 25 63 28 14 14 16 Percent (%) 39.8 34.2 17.4 3.6 24.8 74.5 1.7 11.8 23.0 39.8 .6 3.1 13.0 31.7 10.6 2.5 23.0 .6 11.8 6.8 26.1 45.6 20.5 6.8 15.5 39.1 17.4 8.7 8.7 9.9 - 45 - 5. Data Analysis and Results Reliability and validity were firstly measured at this stage using LISREL v8.51. And then the structure model was tested by SPSS v12.0. Table 5-1 shows the descriptive statistics. Table 5-1: Descriptive Statistics Items PCQ1 PCQ2 PCQ3 PCQ4 PCQ5 PCQ6 PP1 PP2 PP3 SU1 SU2 SU3 SU4 SR1 SR2 SR3 SR4 TI1 TI2 TI3 ISP1 ISP2 ISP3 ISP4 PK1 PK2 PK3 PK4 Jobtenure Industenure Mean 4.8820 4.9130 4.7019 4.6957 4.3043 4.3478 5.0188 5.2236 5.6025 5.1925 5.2050 5.3106 5.1761 2.5528 2.3789 2.6125 2.5000 4.7640 4.8075 4.7081 66.1304 64.9068 54.3727 69.3354 4.5597 4.6289 4.7107 5.0943 2.0818 2.8563 Std. Deviation 1.77616 1.68964 1.75658 1.77145 1.79598 1.70022 1.96661 1.90386 1.60577 1.50630 1.41031 1.37947 1.45626 1.64659 1.52046 1.52953 1.57016 1.76959 1.66777 1.69056 27.85075 28.80469 29.05078 26.94298 1.36220 1.37589 1.37033 1.17351 .86397 1.53306 - 46 - 5.1 Measurement Model Validation of any factor analysis results is essential, particularly when attempting to define underlying structure among the variables. Confirmatory factor analysis by LISREL was therefore conducted for assessing the convergent and discriminant validity of the measurement model. The missing data for main variables were handled by listwise, and the missing data for demographic data were fixed by average. In the end, 154 data were included for examining both measurement model and structural model. Convergent validity was conducted to ensure all items measure a single underlying construct (Bagozzi & Fornell, 1982). The loading of a given item was expected to meet the minimum recommended value of .70. Table 5-2 shows the loadings and t-value of measurement model. PK 4 was loaded at the value of .69 that was very near to the cutoff, so we kept it in the study. However, PP3 was loaded lower than the cutoff. So we deleted this item for validation. Cronbach’s Alpha, composite reliability and average variance extracted (AVE) were also computed for assessing internal consistency of each construct (Hair et al., 1998). Table 5-3 presents the results along these dimensions. All composite reliabilities and Cronbach’s Alpha exceeded the criterion of .70 (Nunnally 1978), while the average variances extracted for these constructs were all above the recommended threshold of .50. The constructs achieved internal consistency. - 47 - Table 5-2: Operationalization of Multiple-Item Sub-Constructs Construct items Perceived Content Quality PCQ1 PCQ2 PCQ3 PCQ4 PCQ5 PCQ6 Physical Proximity PP1 PP2 PP3 Source Understandability SU1 SU2 SU3 SU4 Social Risk SR1 SR2 SR3 SR4 Task Importance TI1 TI2 TI3 Information Source Preference ISP1 ISP2 ISP3 ISP4 Prior Knowledge PK1 PK2 PK3 PK4 Standardized parameter estimate T-value .92 .93 .95 .94 .77 .72 15.20 15.62 16.11 15.90 11.42 10.60 .84 .93 .52 12.05 13.81 6.75 .92 .94 .92 .93 15.22 15.97 15.24 15.62 .76 .87 .88 .93 11.13 13.66 13.84 15.19 .88 .94 .88 14.04 15.60 13.81 .95 .89 .73 .93 16.10 14.49 10.65 15.48 .88 .94 .94 .69 14.01 15.65 15.71 9.8 - 48 - Table 5-3: Assessment of Internal Consistency and Convergent Validity Dimensions Perceived Content Quality Physical Proximity Content Understandability Social Risk Task Importance Prior Knowledge Information Source Preference *Composite reliability = Number of items Cronbach’s Alpha (>.70) 6 2 4 4 3 4 4 .95 .90 .96 .92 .93 .92 .93 Composite reliability* (>.70) .96 .88 .97 .93 .96 .92 .95 (∑ s tan dardized loading ) Average variance extracted** (.50) .80 .71 .90 .78 .88 .87 .83 2 (∑ s tan dardized loading ) + ∑ ε 2 ∑ (s tan dardized loading ) **A VE = ∑ (s tan dardized loading ) + ∑ ε j 2 2 j Where ε = measurement error (1- λ-square), λ = standardized loading Discriminant validity refers to the extent to which the measures for each construct are distinctly different from each other, and is generally checked by comparing the Chisquare of each pair of constructs (Anderson, 1987). Table 5-4 showed the results of 21 pare wise tests. All Chi-square differences were significant at the p[...]... complex task refers to new and genuine decision task in which the paths of task performance and information needed cannot be determined in advance For the former, information need may require task- related information for performing the task For the latter, since the seeker neither knows the paths of task performance nor the task- related information, information need may require all types of information. .. 12 - structure, and objectives The corresponding relationship between information types and information needs was showed in Figure 2-1 Information Types Technical Information Information Needs Task Mastery Referent Information Role Clarification Performance Information Social Integration Social Information Acculturation Organizational Information Figure 2-1: Information Types and Information Needs... is triggered by information needs which are mainly influenced by the task under consideration (e.g., Leckie et al., 1996) Thus, to understand how the task leads to information seeking and consequent information source selection, we firstly review how the task affects information needs which are the cause of information seeking 2.1 Information Needs Information need is the gap between information seekers’... necessary for the seeker to ask every one for information considering cost and time constraints Consequently, the seeker should make a choice among information sources - 13 - In summary, prior research on information seeking proposed three types of task attributes and various types of information sought with respect to information seeking Different information types lead to information source preference of. .. 1961) Therefore, the cost component will be meaningless if quality is not a concern in information seeking Both theoretical and empirical studies have supported that perceived quality is important for information seeking For example, newcomers faced high levels of uncertainty regarding the task role and a sense of acceptance in the groups and organizations, which resulted in information seeking (e.g.,... in information seeking behavior: task complexity, task uncertainty and task importance Task complexity and task uncertainty were proposed to be important features in determining information seeking (e.g., Byström & Järvelin, 1995; Tiamiyu, 1992; Hart - 24 - & Rice, 1991; Culnan, 1983; O’Reilly, 1982; Tushman, 1978) inasmuch as they steered information needs and consequently information seeking For example,... others (e.g., Ashford & Cummings, 1983) In addition to task complexity and task uncertainty, task importance may also be a significant determinant of information need Although no empirical evidence directly suggested that task importance as one of task attributes motivated information seeking, several studies utilized construct that conceptually overlapped task importance provided supportive empirical evidence... translation of a request into “non-natural” language For library, this means locating the call number of the item; for computer-based information, this means use of a command language to formulate a query; information accessibility refers to the ability to physically retrieve the information impendent of any subsequent judgment as to the item’s relevance For computer-based sources, three dimensions of source... and task importance, have been identified in the information seeking behavior literature All three types of task attributes drive an individual to seek information for simplifying task complexity, resolving task uncertainty, or achieving the task In addition, task attributes also varies information needs (e.g., Morrison, 1993a; 1993b; Ashford & Cummings, 1985), which results in various types of information. .. libraries or information collections at your place of work, journals, training sessions, your personal library or files, experts or Forest Service personnel outside of your work unit, and so on Obviously, both information channel and information source were promiscuous Essentially, prior empirical studies had differentiated information channel from information source in the information seeking literature