IEEE TCDL Bulletin
 
space space

TCDL Bulletin
Fall 2007
Volume 3   Issue 3

 

User-adaptive Information Access Supported by Situational Predictors

Applying user's sense-making behavior to adaptivity

Patricia C. Nsc. Souto
PhD Student
Department of Information Science
Loughborough University
Loughborough, Leicestershire, LE11 3TU
<P.C.Nascimento-Souto@lboro.ac.uk>

 

Abstract

Information about a user that supports the adaptation of information (adaptive user model used in adaptive systems) is the key element that enables an effective adaptation decision and result. However, current approaches have been based solely on human dimensions and have excluded other factors and events that drive and influence an individual's information seeking and use behaviors. The current information-needs predictors used to adapt information to users are not sufficient to explain the differences in users' information needs, seeking and use. Consequently, the adaptation of information is impacted, since the results may not reflect what a user needs in a particular situation, in his/her information space and context. The purpose of this research is to include the characteristics of users' problems, or knowledge gaps in the information, that support adaptation. In order to accomplish this, the characteristics of an individual's sense-making processes will be applied as predictors of information needs and behaviors, and as the drivers of information adaptation in adaptive systems such as information management systems or knowledge-based systems in organisational contexts. This proposal changes the focus of the adaptation from being based on users' characteristics to the adaptation being based on the characteristics of users' information needs, seeking and use situations. This research applies a post-constructivist1 approach to adaptive user models and modelling, and the expected result of the research is a 'sense-making-based adaptive framework for information access' that supports an adaptation that accommodates the complex nature of human information needs and behavior, helping individuals in their information-acquisition tasks.

 

1 Introduction

Considering that the core value derived from adapting information2 to a user is the connection between the information accessed and the user's information need,3 the design of the adaptation should consider all aspects surrounding this information need and how it is attended by the user. Thus, in order to make these adaptations based on aspects that are important to users, adaptation strategies and decisions should consider what, where, how and why this information need has emerged and how it has been translated, and equally, should consider how the user seeks4 and uses information to attend his/her need.

The emergence of an information need is influenced by the user's perception and translation of his/her information need, by how the user perceives the situation that caused the information need and by the way he/she performs information seeking and use. Frenette (1999) has pointed out that "humans only use information resources in the context of their own experience – where they come from, what they are struggling with and where they are going". Additionally, Suchman (1987) explained that purposeful actions (i.e., such as information seeking and using) may be planned, but they are "inevitable situated actions", since a user's actions are shaped by the ad hoc and local contingencies of a situation. Thus, it is necessary to look at how actors use existing resources that a particular situation provides.

Information use is situational and contextual,5 and it is influenced by users' internal and external resources. It does not occur independently of the circumstances in which the information use occurs.

Within this context, users have different ways of seeking and obtaining information, and of judging relevance of information under various circumstances. The same types of user may differ in seeking and use information behaviors for identical or different tasks and goals. A single user may also differ in his/her information seeking and use behaviors, either for a single task or for different tasks at different times. The information behavior6 (seeking and use) also differs within and between groups (i.e., the fact that two individuals pertain to the same professional class does not indicate that they will seek and use information in the same way for a given task); within and between domains (i.e., the fact that two different people are researching the same domain does not indicate that they will seek and use information in the same way for a given task); and in relation to tasks (i.e., the fact that two people have identical tasks does not indicate that they will seek and use information in the same way).

Thus, the adaptation of information to a particular user's information needs demands a clear understanding of how the user seeks and uses information and an understanding of what explains the differences and variances in these behaviors. The behavioral 'explanatory' dimensions of the differences in information seeking and use may be identified, modeled and used to effectively drive the adaptation, to help increasing the predictability power of the system and the suitability of adaptation decisions and results.

The adaptation process is directly related to adaptive systems. An adaptive system adapts its behavior to individual users based on information about him/her (it is implicitly collected during the user-system interaction or users are explicitly asked for it), and the adaptive system performs the adaptation using some form of learning, inference or decision making (Jameson, 2003). To be adaptive, the system needs to have enough intelligence to recognize users' differences in seeking and using information. Based on this knowledge, the system infers and identifies the correct procedure to help the user with his/her efforts to solve an information problem. One way to give this intelligence to the system is to have a knowledge base about the user, the so called the user model (UM) (Fig 1).

Flow chart

Figure 1. User modelling or user model application and the adaptation process in adaptive systems (Integrating Brusilovsky, 1996 and Jameson, 2003 user model and user modelling structures)
(For a larger view of Figure 1, click here.)

Thus, the information about the user present in the adaptive UM is a key element in user-adaptivity. It is this information that determines and supports system learning and inferences and subsequently, drives the adaptation of information to the user's needs. According to Torre (2000), the identification of the correct information that attends to the user's needs has been a great challenge in adaptive systems.

In the context of adaptive information access (i.e., search, browse and presentation of search results), the user model and user modelling effectiveness are related to the suitability of the adapted combination of information (documents, articles, papers, summaries, paragraphs, graphics, etc.) to a user's information needs. And from the user's viewpoint, the 'most suitable' information is the information that attends to a specific need that he/she has at a specific moment, and in which he/she has specific resources or deficiencies. It should help the user find, know or improve his/her understanding about something (i.e., not only help to solve problems).

However, as explained above, adaptation in the context of information access, more specifically in the adaptive information retrieval process, has been focused on an atomistic view of the user, often excluding influent factors from his/her information need, information seeking and use behaviors; from the dynamic and interrelated environment in which these behaviors are embedded; from the user's specific information space; and from his/her information problematic situation and context.

This paper explains the proposal for a PhD research project related to the application of an interpretivist and phenomenological theoretical perspective and post-constructivist metatheory to user modelling for adaptivity in an information access context. This research proposes the inclusion of dimensions related to the user's information-seeking and use behaviors, applying the dimensions to an adaptive user model (UM) that aims to drive and support the adaptation of information access.

The remainder of this paper is structured as follows: the second section describes the gaps in current knowledge related to user modelling for adaptive systems, the third section explains how this research aims to fill these gaps, sections 4, 5, and 6 present the research questions, objectives and methodology, respectively.

2 User model in adaptive information access

The focus of this research is the user model that is used7 to support users engaged in an information acquisition process, helping them to find information they need and to access8 the information in the most effective manner. According to Jameson (2003), this function of adaptive systems is comprised of the following sub-functions: (a) helping people find information, (b) tailoring information presentation, and (c) recommending search results. The focus of this research is to help users to find information. According to Brusilovsky & Tasso (2004) and Jameson (2003), this function supports the following processes: hypertext browsing, search or ad-hoc information retrieval, filtering, spontaneous provision of information, and information presentation or visualization of the search results. Within this context, this research focuses only on search or ad-hoc information retrieval and information presentation or visualization.

Commonly, the information in user models includes the following dimensions:

  • the user's preferences, goals, knowledge, background, hyperspace experience, interests, individual traits (cognitive factors, personality factors and learning styles);
  • the user's plans to reach his/her goals;
  • the user's knowledge (beliefs) about a particular domain;
  • the user's general knowledge;
  • usage and environment information (user location and platform);
  • the user's attitudes and proficiencies (e.g., task domain knowledge, proficiency with system);
  • user classification or relevant common characteristics and/or behavior of users pertaining to specific user subgroups (stereotype)
  • the user's perceptual and motor skills limitations, if any (Kobsa, 1990, 1993, 2004, Rich, 1983, Brusilovsky, 1996, p.8, 2001, p.96, 2003, p.1, Kules, 2000, Jameson, 2001).

More recently, affective, emotional and cognitive dimensions have been considered in the area of adaptive systems and computing (Hudlicka & Mcneese, 2002, p.9, Picard & Klein, 2002, Berthouze & Lisetti, 2002, pp. 50-51, Lisetti, 1999).

The user models applied to adaptive information access have mainly adopted the following dimensions of the information about the user to support adaptation:

  • the user's characteristics (i.e., socio demographic and psychographic) or user group characteristics (stereotypes),
  • the user's preferences,
  • the user's interests in topics explicitly manifested by him/her,
  • the user's interests in information inferred mainly by queries terms/keywords/expressions (and history), by the words and by the meaning of words present in titles, summaries and snippets of documents accessed and navigated, by the user's browsing behavior and by interactive actions such as 'save', 'print' or 'bookmark',
  • the user's domain expertise,
  • the user's level of knowledge and experience related to topics and tasks,
  • the user's relevance perception about a particular piece of information (by relevance feedback);
  • the user's cognitive and life style,
  • other's interests in the information (by social navigation and annotation).

To exemplify the information dimensions used in UM that are cited above, some of the adaptive systems applied to information access are outlined below.

Psarras & Jose (2006) designed and developed an adaptive information retrieval system, called PIA (Personal Information Assistant), that adapts document delivery to the multiple and evolving search interests of users. The user model is composed of the user's interests, which are inferred by the keywords used in search iterations. These keywords are derived by extracting terms from the documents, snippets and summaries of documents recently viewed, and weights are assigned to the keywords.

The SiteIF system (Magnini & Strapparava, 2004) provides relevant documents based on the user's interests. These interests are identified by the meanings of the words present in the documents within which the user navigates. The system learns the users' interests by considering the users' browsing activities (requested pages) and thus, it predicts the documents that will be interesting to the user. During a user's browsing in a navigation session, the system creates the user model "as a semantic network whose nodes represent senses (not just words) of the documents requested by the user" (Magnini & Strapparava, 2004, p. 240). Hence, the user model in the SiteIF system is based on users' interests, which are identified by the semantic representation of the browsed information (i.e., news), rather than words. This user model drives the retrieval of new documents with semantic relevance.

Also related to news and their adaptation to users, Ardissono, et al. (2001) have explained the SeAN (Server for Adaptive News) system, which is an adaptive system that enables personalized access to news servers on the Web. Adopting a user model based on stereotypes, SeAN decides which news sections and news items are to be delivered to the user, the level of detail at which each news item should be presented and which advertisements should be shown to a specific user. The user model is initiated with stereotypes and updated by tracking the user's interaction while browsing the news. The user's characteristics and interests that compose the user model are the following: (a) interests, which are the user's interests on topics of news, (b) domain expertise, which is the expertise of user on the topics, (c) cognitive characteristics or cognitive capabilities, and (d) life styles, which are psychographic and socio-demographic data.

Gauch et al. (2003) discussed the Obiwan system, which uses ontologies to provide personalized information access. The user model adopted in this system is automatically and implicitly created, and this process is based on ontology. The user model is comprised of the user's interests by analyzing the user's browsing behavior (content, length, and time spent on each Web page they visit), and no user feedback is used. Each page visited is automatically classified into the concepts of the ontology, attributing and accumulating weights to them.

The Knowledge Sea II project, as explained by Brusilovsky et al. (2005), developed a system that enables personalized information access (ad-hoc information retrieval, information filtering, hypertext browsing, and information visualization), and that helps users to locate resources according to their goals, knowledge and interests. The approach used in this system is known as social navigation, which is heavily based on the feedback of the members of a group. In the system, user navigation is eased and supported by traffic-based visual cues and annotation-based cues. These cues express the density of traffic in a specific topic (shows the volume of people who are using a resource) and the density of annotations made by users. The retrieval of information is based only on the page relevance to the query, and it does not consider the 'social relevance' of links indicated by the traffic navigation cues. The search results are calculated in the traditional way, and only the traffic-based visual cues are shown to users, but the search results do not adapt to these cues.

Mizzaro & Tasso (2002a, 2002b) implemented the Information Retrieval Assistant (IRA), a system that provides suggestions to users who are searching for papers and e-prints in a scholarly publishing portal for physics (the Torii portal). The user model is built by monitoring a user's session history, and any action performed by the user is considered in the building of a session model. The new situations are derived from the history of the user's whole search session.

Ruvini (2003) implemented Toogle, which is a front-end to the Google search engine. Toogle adapts the search results according to the user's browsing behavior over the Google results. As the user goes through the proposed list of results from Google, machine learning techniques infer his/her search goals from clickstream data. The user model is comprised of the user's search goals, which are inferred by the browsing behavior in the list of the search results. This user model is used, then, to reorder the list of results of pages that the user has not yet visited.

Diaz et al. (2005) have developed a Web information system that personalizes summaries of content to users. This personalization occurs according to the combination of short- and long-term user models. The former is created by feedback terms acquired from the system-user interaction, and the latter is created by keywords provided by the user.

Hynecos is an adaptive medical information system that provides hospital information to doctors, nurses, patients and students. According to Vassileva (1994), this system uses individual user models based on the user's level of experience related to a specific task, and its user model drives the adaptation of "the style of viewing, the form of presentation, and the screen layout". The user model is related to a user class model, which is based on profession (e.g., doctor, nurse or patient), location (e.g., ambulance), and ranks within professions (five stages according to the profession).

Waern (2004) focused on adaptive information filtering based on users' profiles, generated automatically (based on feedback) and/or by direct user involvement. The personal profile or user model describes users' interests on a list of topics.

Icarelli & Sciarrone (2004) have developed an information retrieval system that works as an adaptive interface to the search engine Alta Vista™, and which is comprised of two modules: a user modelling system (called HUMOS), and a content-based information filtering (called WIFS). The user modelling system builds and maintains long-term individual user models, which are based on stereotypes. The information filtering module selects html/text documents in the field of Computer Science from the Web, based on the user's interests. The information about the user that is considered in this user model is composed of the user's interests in specific topics within domains. These interests are the user's subjects of interest to which the user assigns weights.

Still within the context of adaptive information retrieval, Mathé & Chen (1996), who developed the seminal research in this field, developed the Adaptive Hyperman system. This system performs adaptivity and adaptability, and it adopts a user model based on the so called 'Adaptive Relevance Feedback' technique. 'Adaptive Relevance Feedback' supports adaptive information retrieval, presentation and navigation, based on a second level of indexing. This second level of indexing is done by users, who assign concepts to information units, and the system automatically builds a conceptual network structure that is used for adaptive information access. The Adaptive HyperMan system enables users to annotate and create hyperlinks within the content, which are judged by other users as relevant or not. These annotations and relevance judgements update and adjust the user model (Rabinowitz et al. 1995).

Related to the adaptive information browsing, Lieberman (2001, 1997, and 1995) developed Letizia, a system that suggests links and recommends web pages to the user according to the page the user is currently browsing. This recommendation is made only according to the current page because "users have many interests over time". The information about the user that is the basis for the adaptation is a compound of the user's interests derived from the recording and analysis of his/her browsing activity in real time. Also developed by MIT, Powerscout (Lieberman et al., 2001) learns the user's preferences by observing and monitoring the user's browsing and then provides recommendations in real time, "but making use of a traditional search engine to search the Web in general".

Also related to adaptive web browsers, Chan's (2000) approach to this topic is creating a user model based on the user's browsing and the content of the pages visited by him/her. A metric is used to identify the user's level of interest, and the user's access patterns are summarized. Then, this metric is used as the basis for the ordering of the search results and page recommendations. The recommendation is also based on collaborative filtering. The sources of information from which to infer user's interests are acquired by monitoring the user's bookmaking actions, the history of the user's requests while in the sessions, access logs and the content of pages visited (links visited) by the user.

Within the function of adaptive information visualization, Leuski & Allan (2004) approached the adaptation of information retrieval based on the users' relevance feedback, and they developed a system called Lighthouse. The user model applied for this adaptation is comprised of the user's assignments of relevance to the documents that were retrieved. The user's attribution of relevance is given to the retrieved documents, and then this attribution modifies and adapts what the authors call the 'User Relevance Model' (URM). The documents considered as relevant are continuously used to adjust the query and to reorder the documents by the "probability of being relevant to the new query". Three types of URMs help the adaptation and visualization of retrieval results by performing the following tasks:

  • ranking documents according to their similarity to other documents that have been previously indicated as relevant,
  • clustering them, and
  • creating an "accurate visual representation of inter-document similarities" of the retrieved set, by mapping the documents by similarity, color coding and relevancy.

Related to adaptive hypertext, Milosavljevic & Oberlander (1998) presented the PEBA II and ILEX systems, which automatically adapt hypertext using techniques from natural language generation. In the ILEX system, the hypertext pages are created dynamically. Thus, they are tailored, in real time, to a single user in a specific communicative situation.9 This adaptation is based on the user's knowledge related to an object (e.g., from a museum gallery), and it is also based on the discourse history or "the objects which the visitor has already seen"9. The PEBA-II system is like an intelligent online encyclopedia, and the hypertext is constructed based on a simple user model, which differentiates between naïve and expert users. This user model also considers the discourse history (Dale et al., 1998, Milosavljevic et al. 1998).

As explained above, current approaches to adaptive information access have excluded some of the factors that give rise to the user's information needs and that largely influence and drive the user's information seeking and use. These factors may include the following:

  • the context of the information need (i.e. the influences of work environment factors, and social-cultural, political and physical contexts; this is not related to the semantic/linguistic context);
  • the role of the individual in relation to the task, context and to the performance;
  • the personal, interpersonal and environmental barriers;
  • and psychological, cognitive, affective needs (based on Wilson's information behavior model, Wilson, 2005, pp. 31-36).

The current UM approaches have also been tentative with regard to including the user's context in the user model. This context was supposed to be represented by the history of the query, snippets and summaries accessed by the user, and consequent inference of user's goals. However, the meaning of 'context' is still ambiguous in user modelling, and in adaptive systems research and literature. Depending on how the term is used, the meaning of 'context' has varied, such as relating to the device employed or to the technical environment in which the user is acting.

The dimensions considered in these UMs are not qualitatively sufficient to explain the differences and variances between users with the same interest. For example, why do two teenagers who need to know about health issues seek and use information completely differently from each another? Why do users from the stereotype 'expert in the domain' seek and use information about quantum physics differently? Even within a domain, or in a group of users (the so called stereotypes), the way each one seeks and uses information can differ. Cheuk (1999) has pointed out the existence of a more dynamic approach that contrasts with the 'domain' approach, whereby "researchers propose that even within a domain, there are differences in people's information seeking and use behaviors at different times, usually construed as in different situations and/or different contexts".

At the root, the dimensions used in current UMs for adaptive information access are derived from a set of concepts predominant in the Computer Science paradigms. The concept of information needs adopted in current UM considers the user as a unidimensional entity, existing statically across time. The conceptual framework on which this perspective is focused, considers the concept of information as an external entity of the individual, whose behavior and complexity are not considered. This perspective is based on findings about users' characteristics that are constant across time and space (e.g., demographic and personality measures) or on characteristics of situations that are also considered as constant (e.g., domain and task characteristics) (Dervin, 2005). However, predicting information seeking and use based only on the most common information about the user as explained above (e.g., socio demographical, psychological and geographical), provides the prediction of differences in information source use (Dervin, 1989, p.56).

Therefore, adaptation of information should also attempt to include other dimensions that can explain the differences in users' informational behaviors. The information about the user included in an adaptive UM should go beyond the user's characteristics, interests and preferences and should consider a more holistic view of the user's information seeking and use behaviors and the complete set of facts and events surrounding these behaviors. In addition, in the information access domain Cole et al. (2005, p. 15) have pointed out that rather than considering that an information need is represented by a query, user-oriented researchers should consider that "a more realistic approach is to bring the user's problem or problematic-situation into the interaction".

The limitations regarding assumptions about user information required for adaptivity have been argued by (Hudlicka & Mcneese, 2002 p.9), who emphasized that the user states (i.e., cognitive, physical, affective and those related to personality) should also be considered and addressed for user modelling and adaptation. Lisetti (1999, pp. 50-51) has also discussed this topic, as follows:
"Researchers in User-Modelling (UM), Human-Computer Interaction (HCI) and Artificial Intelligence (AI) need to learn more about the unsuspected strong interface between affect and cognition. Affective states play an important role in many aspects of the activities we find ourselves involved in, not excluding tasks performed in front of a computer".

Therefore, how the information should be adapted to the user, should consider information beyond the user's characteristics and should also include dimensions related to the situated processes of information seeking and use. These processes are related to specific situations when a gap, or an inadequacy in an user's knowledge is faced. This gap or inadequacy (i.e., not only a problem to be solved) obstructs the user's movement or limits the process of achieving a goal, understanding a topic, solving a problem, creating meaning or completing a task. Since information needs and the information seeking and use processes are not linked solely to the user's characteristics, but also to situations which trigger the seeking and use of information and to the aspects surrounding these processes, the adaptive user models should also consider these aspects and characteristics. Thus, information behavior that is linked to situations and to cognitive events is better predicted when the characteristics of these situations and events are also used, instead of only considering the characteristics of the users (Wilson et al., 2000, p.57). Therefore, the sense-making situations, the cognitive events and the descriptors of context, situations and information seeking and use practices can be added in the user model, improving the predictive power of the UM.

3 Research proposal

Considering the context outlined above, the research discussed in this paper proposes the inclusion of other dimensions or measures of information about the user in adaptive UMs for information access. These dimensions are derived from informational behaviors, specifically from the gap-facing situations in the sense-making process, as perceived by users.

The UM dimensions proposed by this research are called 'situational dimensions'. These situational dimensions will be derived from the user's sense-making situation or the so called sense-making phenomena,10, which is explained by Sense-Making Methodology (i.e., Sense-Making is not a model, but a way of studying information seeking and use).

This meaning of 'situation' is given and supported by the Sense-Making Methodology. In this sense, the concept of 'situation' is considered here as the intersection between the specific situation where the information needs consciously or unconsciously emerged, the gaps that blocked the user from continuing his/her understanding or performance, and the utility that the user hoped to gain from possible answers (i.e., using the information) (see Fig 2).

The 'situational dimensions' will generate the 'situational predictors' of information seeking and use behaviors, which are being proposed to be included in an adaptive UM to support adaptive information access. Thus, by situational predictors, this research means that they are related to the user's sense-making behavior or process, according to Sense-Making Methodology.

Diagram showing conceptualizaton of information need in Sense Making Theory

Figure 2. How 'information need' is conceptualized in Sense-Making Methodology, and the Sense-Making metaphor that is the basis for this concept.
(For a larger view of Figure 2, click here.)

Therefore, this research applies a constructivist approach (i.e., a post-constructivist approach as defined by Brenda Dervin, 19991) to user models for adaptive systems, and this proposal changes the focus of the adaptation from being on the user's characteristics and on the 'context', to being focused on the characteristics of the information need, seeking and use situations/processes, and to the relationship of user and context.

Within this context, the situational predictors explain why identical users (group) or even the same person exhibits different ways to obtain information at different times to complete a single task or similar tasks and why they change them. In a given situation, a user in his/her process of seeking information can have different levels of background knowledge, system experience, affective reactions, cognitive internal and external resources, and perception of the problem to be solved or of the gap that he/she may be facing. A researcher can need to find a focus for a topic in a given situation and then need to confirm ideas and inferences in another situation, both related to the same topic. The same researcher facing another topic may need just to clarify his/her doubts or to choose between two options. Two identical researchers with the same task, goal, background and characteristics, and facing the situation of having to find information to confirm a hypothesis, may have different information needs, seeking and use behaviors.

At the individual level, users with the same demographic characteristics have more probability of sharing certain situational conditions, but it is less probable that they will "actually process and make sense of information in the same way, because they see and define their situations very differently" (Dervin, 1989, p.56].

Some of the Sense-Making studies of information seeking demonstrate that situational characteristics explain far more variance in information seeking and use than do the demographic descriptors of individuals. These Sense-Making studies include the following: Cheuk & Dervin, 1999; Dervin & Shields, 1999; Nilan & Dervin, 1999; Atwood & Dervin, 1982; Dervin, Jacobson & Nilan, 1982; Dervin, Nilan & Jacobson, 1981; and Dervin, Harlock, Atwood, & Garzona, 1980.

The user model with the situational predictors proposed by this research is called 'situational user model' and it will be comprised of the most effective types of information seeking and use predictors (i.e., identified in the data collection and analysis). This 'situational user model' will be incorporated in a corporate information management system or a knowledge-based system (i.e., in an organisational environment). To the 'situational user model', an adaptation model with a set of adaptation rules will be added, compounding the 'sense-making-based and semantic framework for adaptive information access', which is the main outcome of this research. The adaptive models cited above will be represented with semantic web technologies (specifically, ontologies).

The main contributions of this research are the inclusion of a more interpretivist perspective to system-user adaptation, and also the addressing and accommodation of process-and-behavior-based dimensions in an adaptive UM for information access, dimensions that are actually connected to what explains the main differences in information seeking and use behaviors. Thus, the contributions may be evidenced in Information Science and Computer Science and in their research fields such as user modelling, adaptive systems, user-centred design, human-computer interaction, adaptive information retrieval and interactive information retrieval.

4 Research Questions

The main research questions are as follows:

  • How can adaptivity strategies and decisions in the context of information access consider, address and accommodate how individuals really differ and vary in their information needs, seeking and use behaviors? (This question will be addressed in phases 1 and 2 of the research methodology, see Fig. 3.)
  • How can information seeking and use predictors identified by the use of Sense-Making Methodology (situational predictors) be most effectively modeled (in an exclusively or combined way with traditional measures such as demographic characteristics – e.g., age, education level, gender and year of studies), in order to enable, drive and support the adaptation of information? (This question will be addressed in phase 3 of the research methodology, see Fig. 3.)
  • How can the information needs, seeking and use predictors be used for an adaptive user model and user modelling, in order to enable, drive and support flexible, dynamic, responsive and effective adaptation of information as judged by the user's perception and assessment? (This question will be addressed in phases 3 and 4 of the research methodology, see Fig. 3.)
  • How should the user model with situational dimensions be acquired and applied to predict and decide on an adaptation of information most connected to the user's information need, and enables the accommodation of changes on those needs ('information need' as defined in Sense-Making Methodology)? (This question will be addressed in phases 3, 4 and 5 of the research methodology, see Fig. 3.)

5 Research Objectives

5.1 General objective

The general objective of this research is to maximize the predictive power of the information about the user present in adaptive-user models for the information access domain, by applying a post-constructivist approach and proving the value of such an approach by using situational or sense-making measures as predictors of information needs, seeking and use, improving and maximizing the probability of the adapted information being closer to the user's specific information need.

5.2 Specific objectives

  • Identify the situational or sense-making dimensions that most effectively work as predictors of information needs and behavior.
  • Identify the best combination of the traditional and the situational or sense-making dimensions that most effectively work as predictors of information needs and behavior.
  • Model the situational or sense-making predictors (in an exclusive or combined way with the traditional predictors) and propose the 'Situational User Model'.
  • Apply and verify the potential effectiveness of the adaptation decisions (e.g., does the adapted information 'answer' the 'questions' in a specific information seeking situation and help the gap-bridging? Does the adapted information attend to the 'uses/helps' expected by users?11).
  • Systematize the user model and the user modelling rules in an adaptive framework for information access (i.e., the 'sense-making-based and semantic framework for adaptive information access').

6 Research Methodology

This study will adopt the Sense-Making assumptions, theoretic perspective, methodological approach and research methods for studying and modelling information needs, seeking and use, and consequently, will generate the dimensions that will be included in the adaptive user model.

The Sense-Making Methodology enables the study of the sense-making processes of individuals and applies a coherent and consistent theory of how to study these processes. The Sense-Making Methodology is a consolidated methodology (30 years, more than 100 researches, according to Dervin, 2001, p.61) and has been used in many studies about prediction and modelling of information needs and behavior (which is a critical issue for this research).

Additionally, at the same time Sense-Making Methodology offers a qualitative approach for interviewing and data collecting – a quantitative approach because it developed procedures and a structure for data analysis and systematization (i.e., the sense-making metaphor). And the Sense-Making Methodology is systematic, because the interviewing, listening, understanding and systematization of this user understanding is guided by a theory (Dervin, 1989).

Considering this, the research comprises the five phases illustrated in Fig. 3. Phases 2, 3, 4 and 5 will occur iteratively and will be performed in parallel. All the phases in data collection and data analyses are explained in the following sections.

Chart showing research phases

Figure 3. Overview of this research phases.
(For a larger view of Figure 3, click here.)

6.1 Hypothesis

  • PHASES 1 AND 2:
  • Situational measures are more effective in driving and supporting the prediction of information seeking and use behaviors and aspects, either in an exclusive or a combined way with the traditional measures (since the situational measures make up the largest part of the set of predictors).
  • PHASE 5:
  • The adaptation of information based on the use of situational predictors in the adaptive user-model is more effective in helping a user to overcome his/her gaps of knowledge (i.e., in gap-bridging) in a specific situation.

6.1 Data Collection

6.1.1 Sample selection

For the data collection in phase 1, the sample will be composed of 12 to 20 participants. The participants sample will be composed by using a purposeful sampling (nonrandom sampling), according to specific characteristics related to the phenomena which will be investigated.

The participants are knowledge workers (i.e. corporate heavy-users of information) (e.g., analysts, consultants, managers, lawyers and government professionals) of information-intensive organizations12 (e.g., education, health, research & development and telecommunications), who routinely and intensively use information as a work practice and for the purpose of knowledge creation13 (e.g. planning, new products development and marketing functions). The participants should be recruited from different corporations and should be comprised of different kind of professions and functions.

6.1.2 Data collection

The data collection will be collected in two separate phases: in Phase 1 and Phase 5 (see Fig. 3).

  • PHASE 1:
  • (a) Identification of structural or traditional measures (demography dimensions (e.g., age, education level, gender, year of studies, income, computer ownership and nationality), information literacy levels and domain knowledge.
  • A web-based questionnaire will be used.
  • (b) Identification of situational or alternative measures (derived from sense-making) – These measures are related to the description of the sense-making behaviors, and the variables studied in this phase were explained in item 4. They are a compound of the following situations-gaps-uses categories of variables: actor's sense-making situations, gaps in sense-making, actor-defined purpose, information using strategy, information values, information traits.
  • This data will be collected using the Sense-Making Methodology and the Micro-Moment Time-Line interview.
  • PHASE 5:
  • (a) Evaluation of the effectiveness of the situational UM adaptation decisions – evaluation with respect to the prediction of information that should be delivered and presented as a result of information search. This evaluation will be formative and it will be developed iteratively throughout the prototype development. The entire development cycle will be covered by iterative evaluations, and at the final stage, a summative evaluation will also be performed. It is worth noting that it will not be a usability evaluation.
  • The UM evaluation will be based on the 'Decomposition Model for the Layered Evaluation', developed by Paramythis & Weibelzahl (2005).
  • In this phase, the prototype of the situational user model is supposed to be installed in an information management system or knowledge-based system (information use in an organisational context).
  • The data collection in Phase 5 will occur as follows:
  • (i) Data collection pre-interaction with the adaptive system with the 'situational user model' proposed by this research – the goal here is to identify the context-of-use measures or the context in which the user/respondent is seeking information.
  • (ii) Data collection during the interaction with the adaptive system with the 'situational user model' proposed by this research – the objective here is to identify what and how the user interacts with the information retrieved and made available for browsing by the system.
  • (iii) Data collection post-interaction with the adaptive system with the 'situational user model' proposed by this research – the goal here is to evaluate the adapted information delivered with respect to the situation and gaps elicited by users in the beginning.

6.2 Data Analyses

The unit of analysis is the sense-making instance, which is composed by situations, gaps and uses/helps.

The data analyses will be made in phases 2 and 5 (see Fig. 3).

  • PHASE 2
  • (a) Coding and categorization – content analysis for the situational predictors and the situational categories of 'situation-gaps-uses' will be derived deductively using the Sense-Making Methodology (see Fig. 2).
  • (b) Verification of predictive power of traditional and situational or sense-making measures – a specific set of statistical techniques will be used.
  • (c) Establishment of the set of predictors that will be used in the 'situational user model'.
  • PHASE 5
  • The data analysis in Phase 5 will be made iteratively, during and after the development of the representation of the 'situational user model' and of the adaptation rules (both ontology-based), the development of a situational user model prototype and its implementation in an information management system or knowledge-based system.
  • The data analysis will be comprised of the analysis of the effectiveness of the situational UM by assessing the quality of the adaptation decisions.

Notes

1. Post-constructivist refers to the use of the Sense-Making Methodology as a theoretical foundation and also its methodology and methods. Dervin (1999, p. 730) explained that the Sense-Making Methodology was described by her in (Dervin, 1983) as "a constructivist approach, while now I describe it as post-constructivist, or postmodern modernist. In Dervin (1994), I termed sense-making as communitarian, but revised my thinking in Dervin (1998) when I termed it a verbing approach".

2. The "adaptation of information" is related to adaptive systems and to adaptive information access, which functions are to help and support people in finding information (searching, browsing and visualization).

3. Information need is conceptualized here based on the Sense-Making Methodology (see Fig. 2).

4. Information Seeking Behavior is defined as "the purposive seeking for information as a consequence of a need to satisfy some goal. In the course of seeking, the individual may interact with manual information systems (such as a newspaper or a library), or with computer-based systems (such as the World Wide Web)" (Wilson, 2000).

5. The definition of 'context' used here is the one given by McCreadie & Rice (1999, p. 58): "The larger picture in which the potential user operates; the larger picture in which the information system is developed and operates, and potential information exists".

6. Information behavior is defined as "...the totality of human behavior in relation to sources and channels of information, including both active and passive information seeking, and information use. Thus, it includes face-to-face communication with others, as well as the passive reception of information as in, for example, watching TV advertisements, without any intention to act on the information given (Wilson, 2000. Another definition is "the human behavior dealing with generation, communication, use and other activities with information, such as information seeking and behavior and interactive information retrieval" Ingersen & Jarvelin (2005, p. 21).

7. User models have been applied in learning, collaboration, information access and TV programming.

8. Access here is considered as physical access and cognitive access.

9. <http://www.hcrc.ed.ac.uk/ilex/>.

10. 'Sense-Making' with initial upper case refers to the methodology, and 'sense-making' with initials lower case refers to the phenomena.

11. Gaps, gap-bridging, uses-helps and situation are all terms from Sense-Making Methodology.

12. The classification of the information-intensive organisations according to Eurostat was shown in Laafia (2002, p.7).

13. Information use for knowledge creation is one of the three arenas of information use in an organizational context (Choo, 2006).

References

Ardissono, L., Console, L., Torre, I., (2001). An adaptive system for the personalized access to news. AI Communications, v. 14(3), pp. 129-147. Available at <http://www.di.unito.it/~liliana/pubbl.html>.

Atwood, R., & Dervin, B. (1981). Challenges to sociocultural predictors of information seeking: A test of race versus situation movement state. In M. Burgoon (Ed.), Communication yearbook, v. 5, pp. 549-569. New Brunswick, NJ: Transaction Books.

Berthouze, N. B., Lisetti, C. L. (2002). Modelling Multimodal Expression of User's Affective Subjective Experience. User modelling and User-Adapted Interaction, v. 12(1), pp. 49-84.

Brusilovsky, P. (1996). Methods and techniques of adaptive hypermedia. User modelling and User-Adapted Interaction, v. 6 (2-3), pp. 87-129.

Brusilovsky, P. (2003). Developing adaptive educational hypermedia systems: from design models to authoring tools. In: T. Murray, S. Blessing and S. Ainsworth (eds.): Authoring Tools for Advanced Technology Learning Environment, Dordrecht: Kluwer Academic Publishers, pp. 377-409.

Brusilovsky, P., Farzan, R., and Ahn, J., (2005). Comprehensive personalized information access in an educational digital library. In: Proceedings of The 5th ACM/IEEE-CS Joint Conference on Digital Libraries, Denver, CO, June 7-11, 2005, ACM Press, pp. 9-18.

Brusilovsky, P., Tasso, C., 2004. Preface to Special Issue on User Modeling for Web Information Retrieval - User Modeling and User-Adapted Interaction 14: 147-157.

Chan, P. (2000). Constructing Web User Profiles: A Non-Invasive Learning Approach. In Web Usage Analysis and User Profiling, LNAI 1836, Springer-Verlag, 2000, pp. 39-55.

Chen, L., Sycara, K. (1998). A Personal Agent for Browsing and Searching. In Proceedings of the 2nd International Conference on Autonomous Agents, 1998, pp. 132-139.

Cheuk, B. W-Y., & Dervin, B. (1999). A qualitative sense-making study of the information seeking situations faced by professionals in their workplace contexts. Electronic Journal of Communication, v. 9(4).

Chin, D., 2003. Evaluating the Effectiveness of User Models by Experiments. In: 9th International Conference on User Modeling (UM'2003) Workshop Proceedings, 22 to 26 of June 2003, at the University of Pittsburgh, Johnstown, PA, USA. Available online at <http://www2.ics.hawaii.edu/~chin/UM/um03tutorial.html>.

Choo, Chun Wei. (2006). The Knowing Organization: How Organizations Use Information to Construct Meaning, Create Knowledge, and Make Decisions. 2nd edition. New York: Oxford University Press.

Cole, C., Beheshti, J., Leide, J. E., Large, A., 2005. Interactive information retrieval: bringing the user to a selection state. In: Spink, A., Cole, C., New directions in cognitive information retrieval, Springer, Dordrecht, The Netherlands, 2005, 13-41.

Dale, R., Oberlander, J., Milosavljevic, M. Knott, A., (1998). Integrating Natural Language Generation and Hypertext to Produce Dynamic Documents. Interacting with Computers, v. 11(2), 109-135.

Dervin B. (1994). Information <----> democracy: An examination of underlying assumptions. Journal of the American Society for Information Science, 45 (6), 369-385.

Dervin B. (1999a). On studying information seeking methodologically: The implications of connecting metatheory to method. Information Processing and Management, 35, 727-750.

Dervin, (1983) - Dervin, B. 1983, An overview of Sense-Making research: Concepts, methods, and results to date. Paper presented at the annual meeting of the International Communication Association, Dallas, TX.

Dervin, B. (1998). Sense-Making theory and practice: An overview of user interests in knowledge seeking and use. Journal of Knowledge Management, 2 (2), 36-46.

Dervin, B. (1989). Users as research inventions: how research categories perpetuate inequities. In: Dervin, B., Foreman-Wernet, L. (Eds.) (2003). Sense-Making Methodology reader: Selected writings of Brenda Dervin. Cresskill, NJ: Hampton Press, pp. 215-231.

Dervin, B. (2005). What methodology does to theory: Sense-Making Methodology as exemplar. In Karen Fisher, Sanda Erdelez, Lynne (E.F.) McKechnie (Eds.), Theories of information behavior, pp. 25-30. Medford, NJ: Information Today (2005).

Dervin, B. (2001). What we know about information seeking and use and how research discourse community makes a difference in our knowing. Background paper prepared for Health Information Programs Development, National Library of Medicine, Bethesda, MD. Available [online]: <http://communication.sbs.ohio-state.edu/sense-making/art/artabsdervin01nlm.html>.

Dervin, B., Harlock, S., Atwood, S., & Garzona, C. (1980). The human side of information: An exploration in a health communication context. In D. Nimmo (Ed.), Communication yearbook, v. 4, pp. 591-608. New Brunswick, NJ: Transaction Books.

Dervin, B., Jacobson, T. L., & Nilan, M. S. (1982). Measuring qualitative and relativistic aspects of information seeking: A test of qualitative-quantitative methodology. Paper presented at the annual meeting of the International Communication Association, Boston.

Dervin, B., Nilan, M. S. & Jacobson, T. L. (1981). Improving predictions of information use: A comparison of predictor types in a health communication setting. In D. Nimmo (Ed.), Communication Yearbook, v. 5, pp. 807-830. New Brunswick, NJ: Transaction Books.

Dervin, B., Shields, P. (1999). Adding the missing user to policy discourse: Understanding U.S. user telephone privacy concerns. Telecommunication Policy, v. 23, pp. 405-435.

Diaz, A., Gervas, P., Garcia, A., (2005). Evaluation of a system for personalized summarization of Web contents. In Ardissono, L., Brna, P., Mitrovic A. (Eds.): UM 2005, 10th. International Conference, Edinburgh, Scotland, UK, July 2005, pp. 453-462.

Frenette, M., (1999). Explorations in adolescents' sense-making of anti-smoking messages. Electronic Journal of Communication, v. 9(2).

Gauch, S., Chaffee, J., Pretschner, A., (2003). Ontology-Based User Profiles for Search and Browsing. User modelling and User-Adapted Interaction, Special Issue on User modelling for Web and Hypermedia Information Retrieval.

Hudlicka, E., Mcneese, M., (2002). Assessment of User Affective and Belief States for Interface Adaptation: Application to an Air Force Pilot Task. User modelling and User-Adapted Interaction, v. 12(1), pp. 1-47. Kluwer Academic Publishers, Netherlands.

Icarelli, A., Sciarrone, F., (2004). Anatomy and Empirical Evaluation of an Adaptive Web-Based Information Filtering System. User modelling and User-Adapted Interaction, v. 14, pp. 159-200.

Ingersen, P., Jarvelin, K., 2005. The turn: integration of information seeking and retrieval in context, Springer, The Netherlands.

Jameson, A. (2001). User-Adaptive and Other Smart Adaptive Systems: Possible Synergies. Proceedings of the First EUNITE Symposium, Tenerife, 13-14 December 2001.

Jameson, A. (2003). Adaptive Interfaces and Agents. In J. Jacko & A. Sears (Eds.), Human-computer interaction handbook, pp. 305–330, Mahwah, NJ: Erlbaum.

Kelly, D. & Teevan, J. (2003). Implicit feedback for inferring user preference: A bibliography. In SIGIR Forum Proceedings, v. 37(2), pp. 18-28.

Kobsa, A. (1990). User modelling in dialog systems: Potentials and hazards. Technical Report Bericht Nr. 64, Universitat des Saarlandes, Informatik, Saarbrucken. Sonderforschungsbereich 314. Kunstliche Intelligenz, Wissensbasierte Systeme.

Kobsa, A. (1993). User modelling: Recent Work, Prospects and Hazards. In Schneider-Hufschmidt, T. Khme, U. Malinowski (Eds.) Adaptive User Interfaces: Principles and Practices.

Kobsa, A. (2004). Adaptive Interfaces. In: W. S. Bainbridge, ed.: Encyclopedia of Human-Computer Interaction. Great Barrington, MA: Berkshire Publishing.

Kules, B. (2000). User modelling for Adaptive and Adaptable Software Systems. Available at <wmk@takoma-software.com>. Accessed on May 2006.

Laafia, I. (2002). Employment in High Tech and Knowledge Intensive Sectors in the EU Continued to Grow in 2001. Statistics in Focus: Science and Technology, Theme, 9(4).

Leuski, A., Allan, J., (2004). Interactive information retrieval using clustering and spatial proximity. User modelling and User-Adapted Interaction, v. 14, pp. 259-288.

Lieberman, H. (1995). Letizia: An Agent That Assists Web Browsing. In Proceedings of the 14th International Joint Conference on Artificial Intelligence, 1995, pp. 924-929.

Lieberman, H. (1997). Autonomous Interface Agents. In Proceedings of the ACM Conference on Computers and Human Interaction (CHI'97), May 1997.

Lieberman, H., et al. (2001). Why Surf Alone? Exploring the Web with Reconnaissance Agents", in Communications of the ACM, pp. 69-75.

Lieberman, H., Fry, C. and Weitzman, L. (2001). Exploring the Web with Reconnaissance Agents. Communications of the ACM, v. 44(7), pp. 69-75.

Lisetti, C. (1999). Modelling cognition-emotion of users for improved interaction with software systems. In Proceedings of the seventh international conference on user modelling, Banff, Canada, pp.307-309.

Magnini, B., Strapparava, C., 2004. User modelling for News Web Sites with Word Sense Based Techniques. User modelling and User-Adapted Interaction, v.14, pp. 239-257.

Martinez, E. F., Magoulas, G., Chen, S.; Macredie, R., (2005). Modelling human behavior in user-adaptive systems: Recent advances using soft computing techniques. Expert Systems with Applications, v. 29 (2), pp. 320-329.

Mathé, N. & Chen, J. R., 1996. User-Centered Indexing for Adaptive Information Access. User Modeling and User-Adapted Interaction, 6, pp. 225-261.

McCreadie, M., Rice, R. E., 1999. Trends in Analyzing Access to Information. Part I: Cross-Disciplinary Conceptualizations of Access. Information Processing & Management, v35 n1 p45-76 Jan 1999.

Milosavljevic, M. Tulloch, A., Dale, R. (1996). Text Generation in a Dynamic Hypertext Environment. In Proceedings of the Nineteenth Australasian Computer Science Conference (ACSC'96), ACSC 18(1), Melbourne, Australia, pp. 301-307.

Mizzaro S., Tasso C., (2002a). Ephemeral and persistent personalization in adaptive information access to scholarly publications on the Web. Artificial Intelligence Laboratory Department of Mathematics and Computer Science.

Mizzaro, S., Tasso C., (2002b). Personalization techniques in the TIPS Project: The Cognitive Filtering Module and the Information Retrieval Assistant. In Mizzaro, S., Tasso C., (Eds.), Personalization Techniques in Electronic Publishing on the Web: Trends and Perspectives - Proceedings of the AH2002 Workshop, Universidad de Màlaga, Spain, May 2002, pp. 89-93.

Nilan, M. S., & Dervin, B., (1999). Beyond agency to structure: Moving quantitative sense-making studies to a focus on both societal structural arrangements and information seeking agency. Electronic Journal of Communication, 9(4).

Oppermann, R., (1994). Adaptively supported adaptability. International Journal of Human Computer Studies, 40(3), 455-472.

Paramythis, A., Weibelzahl, S., 2005. A Decomposition Model for the Layered Evaluation of Interactive Adaptive Systems. In: Ardissono, P., Brna, P., Mitorovic, A., (Eds.), User Modeling 2005 Proceedings, pp. 438-442.

Picard, R. W., Klein J., (2002). Computers that recognize and respond to user emotion: theoretical and practical implications. Interacting with computers, v.14, pp. 141-169.

Pretschner A., (1999). Ontology based personalized search. Master's thesis, The University of Kansas, Lawrence, KS, 1999.

Psarras, I., Jose, J. A , 2006. System for Adaptive Information Retrieval. In: Wade, V., Ashman, H., Smith, B. (Eds.) Adaptive Hypermedia and Adaptive Web-based Systems Proceedings, Lecture Notes in Computer Science, (2006), 4018, 313-317.

Rabinowitz, J., Mathé, N. & Chen, J. R., (1995). Adaptive HyperMan: A Customizable Hypertext System for Reference Manuals. In Proceedings of the AAAI Fall Symposium on Artificial Intelligence Applications in Knowledge Navigation and Retrieval, Cambridge, MA, Nov. 10-12.

Rich, E., (1999): Users are Individuals: Individualizing User models. In International Journal of Human Computer Studies, 51 (2) pp. 323-338.

Ruvini, Jean-David, (2003). Adapting to the user's internet search strategy. In Brusilovsky et al. (Eds.), UM 2003 User modelling Proceedings 9th International Conference, Johnstown, PA, USA, June, 2003, pp. 55-64.

Suchman, L., 1987. Plans and Situated Actions. Cambridge University Press, Cambridge, UK, 1987.

Torre, I., (2000). A Modular Approach for User modelling. In Proceedings of the International Conference on Adaptive Hypermedia and Adaptive Web-Based Systems, v. 1892, pp. 415-420.

Trajkova, J., Gauch, S. (2004). Improving Ontology-Based User Profiles. In RIAO: Recherche d'Information Assisté;e par Ordinateur (computer assisted information retrieval) Proceedings, University of Avignon (Vaucluse), France, April 26-28.

Vassileva, J., (1994). A practical architecture for user modelling in a hypermedia-based information system. In: 4th International Conference on User modelling, Hyannis, MITRE.

Waern, A., (2004). User Involvement in Automatic Filtering: An Experimental Study. User modelling and User-Adapted Interaction, 14, pp. 201-237.

Willms, Swantje, (2003). Visualizing a user model for educational adaptive information retrieval. In Brusilovsky et al. (Eds.), UM 2003 User modelling Proceedings 9th International Conference, UM 2003, Johnstown, PA, USA, June, 2003, pp. 432-434.

Wilson, T. D., Ford, N. J., Ellis, D., Foster, A. E., Spink, A. (2000). Uncertainty and its correlates. The New Review of Information Behavior Research, v. 1, pp. 69-84.

Wilson, T.D, 1999. Models in information behavior research. Journal of Documentation, v. 55(3), pp. 249-270.

Wilson, T.D. (2005). Evolution in information behavior modelling: Wilson's model. In Karen Fisher, Sanda Erdelez, Lynne (E.F.) McKechnie (Eds.), Theories of information behavior, Medford, NJ: Information Today, pp. 31– 36.

Wilson, T.D. (2000). Human Information Behavior. Informing Science, Special issue on Information Science Research, v. 3(2).

 

© Copyright 2007 Patricia C. Nsc. Souto

Top | Contents
Previous Article
Home | E-mail the Editor