Reader's Guide
Sixth International Conference on User Modeling On-Line Proceedings

This guide is intended to help readers of the UM97 proceedings volume see how the UM97 papers and posters fit into the big picture of user modeling research.

Introduction

There are several questions that can be asked about most instances of user modeling research. One way of bringing to light the relationships among the UM97 papers and posters is to compare the corresponding answers to these questions, in all cases in which the questions are applicable.

The following six sections address the following questions in turn, assuming that in each case considered, user modeling techniques are being investigated that are to enable some system S to adapt to each individual user U.

1. Purpose of user modeling

   In what way is S's adaptation to U intended to be beneficial to U?

2. Content of the user model

   What sort of information about U is represented in S's user model?

3. Methods for exploiting the user model

   According to what principles or inference techniques does S decide how to adapt its behavior on the basis of the information in its user model?

4. Input data for user model acquisition

   On the basis of what types of evidence does S construct its user model?

5. Methods for constructing the user model

   According to what principles or inference techniques does S arrive at the hypotheses about U that are stored in the user model?

6. Empirical foundations

   What sorts of empirical data give us reason to believe that S's methods are valid and useful?

By selecting one of the articles cited below, you can jump to a separate page with its abstract (and other information) that in turn offers links to the full manuscript.

1. Purposes of user modeling

Help U to find information

• Remind U of previous Web navigation paths (Maglio and Barrett)
• Recommend Web pages (Akoulchina and Ganascia)
• Implicitly recommend Web hyperlinks (Gori et al.)
• Recommend potentially suitable existing solutions to a problem (e.g., airplane flights) (Linden et al.)
• Select documents of interest to U (Benaki et al.)
• Adapt hypertext links and recommendations to U's previous navigation behavior (Staff)
• Adapt hypertext navigation mode to U's knowledge (De Carolis and Pizzutilo)
• Filter WWW documents in accordance with U's interests (Ambrosini et al.)

Tailor information presentation to U

• Adapt appearance of charts to U's abilities and preferences (Gutkauf et al.)
• Choose appropriate form of presentation with text and/or diagrams (Kalyuga et al.)
• Take into account U's available working memory capacity (Schäfer and Weyrath)
• Adapt comparisons in text to U's knowledge (Milosavljevic)
• Tailor arguments to U's beliefs and values (Grasso)
• Adapt hypermedia presentations to U's interests and knowledge (Fink et al.; De Carolis and Pizzutilo)
• Adapt health-education documents to individual patients (Hirst et al.)
• Present decision-supporting information according to U's personality and preferences (Paranagama et al.)
• Adapt handling of incorrect answers to the student's preferences (Murphy and McTear)

Adapt an interface to U

• Offer Web navigation shortcuts that reflect past accesses (Maglio and Barrett)
• Facilitate selection of presumably relevant Web hyperlinks (Gori et al.)
• Adapt interface features and hints to U's familiarity with S (Brusilovsky and Schwarz)
• Translate high-level visualization preferences into concrete camera control actions (Bares and Lester)
• Recommend settings for technical devices (Doux et al.)
• Recommend keyboard adaptations for users with disabilities (Trewin and Pain)
• Adapt a hypermedia interface to U's disabilities (Fink et al.)
• Suggest corrections of (idiosyncratic) spelling errors of dyslexic users (Spooner and Edwards)
• Offer a suitable next move after an unexpected, ambiguous dialog act (Stein et al.)
• Provide special support and interface simplifications for novice users (Strachan et al.)

Choose suitable instructional exercises or interventions

• Choose math exercises taking into account U's subskill proficiencies (Beck et al.)
• Generate medical cases of an appropriate difficulty level (Carberry and Clarke)
• Select suitable language exercises (Murphy and McTear)
• Guide U toward suitable instructional Web pages (Weber and Specht)
• Present instructional examples from U's own learning history (Weber and Specht)
• Tailor instructional interventions to U's state of knowledge (Conati et al.)
• Take into account changes in a student's beliefs (Giangrandi and Tasso)
• Derive a differentiated assessment of a trainee's problem solving skills (Moinard and Joab)
• Support mastery learning by tracing the development of U's knowledge (Corbett and Bhatnagar)

Give U feedback about U's knowledge

• Provide to students feedback on their strengths and weaknesses in foreign language writing (Bull)

Support collaboration

• Select appropriate collaborators (or help U to do so) and facilitate communication between collaborators (Collins et al.)
• Recommend specific forms of collaboration between students (Bull and Smith)

Predict U's future behavior

• Predict correct and incorrect answers of a student (Chiu et al.)
• Predict goals, actions, and locations of an agent in a large domain (Albrecht et al.)

[Other functions]

• Verify U's competence to add information to S (Akoulchina and Ganascia)
• Take into account U's cooperativeness, sincerity, and credulity (Quaresma and Lopes)
• Anticipate other agents' actions so as to coordinate with them (Noh and Gmytrasiewicz)
• Enable U to write and debug programs using high-level concepts that U finds natural (Seta et al.)
• Support various types of adaptation with a general user modeling shell system (Pohl and Höhle)
• Take into account factors such as U's relationship with S and the importance of U's goals (Vassileva)

2. Content of the user model

U's preferences, interests, attitudes, and goals

• Preferences concerning possible solutions to a problem (Linden et al.)
• Weights of decision-relevant attributes (Paranagama et al.)
• Payoff matrices that underlie U's behavior (Noh and Gmytrasiewicz)
• User-specific low-level parameters concerning camera control (Bares and Lester)
• Preferences concerning aspects of charts (Gutkauf et al.)
• Preferences concerning the modality of hypermedia-presented information (Fink et al.)
• Preferences and attitudes concerning aspects of language learning (Murphy and McTear)
• Context of current hypertext node, which reflects U's interests (Staff)
• U's interests with respect to information on the WWW (Ambrosini et al.)
• Goals of U's Web searches (Akoulchina and Ganascia)
• General goals motivating U's use of a hypermedia system (Fink et al.)
• Problem solving strategy currently pursued by U (Conati et al.)
• Attitudes concerning medical issues (Hirst et al.; Grasso)
• Cooperativeness, sincerity, and credulity; and specific goals (Quaresma and Lopes)

Specific aspects of U's knowledge and beliefs

• Knowledge concerning features of a complex interface (Brusilovsky and Schwarz)
• Knowledge of particular concepts in instructional material (Weber and Specht)
• A student's strengths and weaknesses in a subject area (qualitatively and quantitatively assessed) (Bull)
• A student's knowledge of particular problem-solving rules (Chiu et al.; Corbett and Bhatnagar; Conati et al.)
• Ability to perform specific steps of a task (Collins et al.)
• Characterization of skills at the operational, tactical, and strategic levels (Moinard and Joab)
• Proneness to and causes of particular language errors (Murphy and McTear)
• Rules that underlie U's incorrect spelling behavior (Spooner and Edwards)
• U's knowledge of domain concepts relevant to a hypermedia presentation (Fink et al.; Milosavljevic)
• U's overall familiarity with the subject matter of a hypertext (De Carolis and Pizzutilo)
• U's factual beliefs about medical issues (Grasso)
• Dialog-relevant factual beliefs (Quaresma and Lopes)
• Beliefs held during particular (underspecified) time intervals (Giangrandi and Tasso)

U's proficiencies

• Mastery of particular math subskills (Beck et al.)
• Command of the declarative knowledge relevant to problem solving (Corbett and Bhatnagar)
• Proficiency with respect to the target language and the domain (Murphy and McTear)
• Level of domain expertise (Kalyuga et al.)
• A medical student's level of diagnostic expertise (Carberry and Clarke)
• Rates at which U learns and forgets instructional content, respectively (Beck et al.)
• Ability to handle particular topics individually and in collaboration (Bull and Smith)
• Competence in dealing with computers and with a specific hypermedia system (Fink et al.)
• Proficiency in the task domain and in the use of S (Strachan et al.)
• U's domain expertise and theoretical orientation (Akoulchina and Ganascia)
• U's familiarity with emergency situations (Schäfer and Weyrath)

U's noncognitive abilities

• Visual perceptual abilities; mental rotation ability (Gutkauf et al.)
• Perceptual and motor abilities relevant to both computer use and real-world activities (Fink et al.)

U's Personal characteristics

• Location, job title, etc., of potential collaborators (Collins et al.)
• Personal characteristics recorded in U's medical record (Hirst et al.)
• Level of education, age, etc. (Murphy and McTear; De Carolis and Pizzutilo)
• Personality type (Paranagama et al.)

History of U's interaction with S

• Aspects of U's keyboard use (Trewin and Pain)
• U's WWW navigation history (Weber and Specht; Gori et al.; Maglio and Barrett)
• History of interaction with hypermedia system (Milosavljevic)
• History of dialog acts (Stein et al.)
• U's use of interface features and reading of hints about them (Brusilovsky and Schwarz)
• U's execution of specific steps in the current task (Collins et al.)
• Observed actions and locations of U within a large domain (Albrecht et al.)

[Other types of content]

• Assignment to one of a set of classes of similar users of technical devices (Doux et al.)
• Available working memory capacity, emotional state, etc. (Schäfer and Weyrath)
• U's higher-order beliefs about the system's payoffs and beliefs (Noh and Gmytrasiewicz)
• U's goal priorities, emotions, moods, and relationship with S (Vassileva)
• A task ontology that is suited to U's way of thinking (Seta et al.)

3. Methods for exploiting the user model

Decision-theoretic methods

• Quantitative evaluation of possible solutions according to U's preferences (Paranagama et al.; Linden et al.)
• Recursive Modeling Method for predicting another agent's actions (Noh and Gmytrasiewicz)

Logic-based techniques

• Abduction (Stein et al.)
• Logic programming (Quaresma and Lopes)
• Various inference techniques within a modal logic framework (Pohl and Höhle)

Bayesian methods

• Probabilistic prediction of rule mastery on the basis of past performance and level of declarative knowledge (Corbett and Bhatnagar)
• Use of Bayesian networks to predict a student's problem solving behavior (Conati et al.)
• Dynamic Bayesian Networks for prediction of temporally variable actions and properties of U (Schäfer and Weyrath; Albrecht et al.)

Machine learning techniques

• Use of Input-Output Agent Modeling to predict a student's responses (Chiu et al.)
• Use of neural networks to predict U's interest in Web pages (Gori et al.)
• Use of K-Means classification technique to find a behavior close to the one that U would choose (Doux et al.)

Other general techniques and principles

• General techniques for the sequencing of instructional material (Brusilovsky and Schwarz)
• General search techniques and heuristics (Spooner and Edwards)
• Use of semantic networks to assess the relevance of documents to U's interests (Ambrosini et al.)
• Episodic learner modeling for retrieval of suitable instructional examples (Weber and Specht)
• Formalization of rhetorical techniques (Grasso)
• Hypertext architecture in which context is taken into account (Staff)
• Techniques for executing and tracing conceptual-level programs (Seta et al.)

Application-specific computational procedures

• Computations concerning potentially interesting Web pages (Akoulchina and Ganascia)
• Algebraic technique for choosing suitable math exercises (Beck et al.)
• Quantitative criteria for determining keyboard adaptation recommendations (Trewin and Pain)

Application-specific qualitative rules and procedures

• Method for processing a history of Web page visits (Maglio and Barrett)
• Criteria for recommending the next instructional Web page to visit (Weber and Specht)
• Rules for providing a simpler interface and more support to novice users (Strachan et al.)
• Rules for adapting hypermedia presentations to various properties of users (Fink et al.)
• Rules linking user properties with hypertext generation parameters (De Carolis and Pizzutilo)
• Hypertext search techniques that take context into account (Staff)
• Rules for taking into account preferences and abilities relevant to chart design (Gutkauf et al.)
• Provision for queries to U's medical record in an authoring environment (Hirst et al.)
• Rules based on empirically determined relationships between domain expertise and appropriate presentation format (Kalyuga et al.)
• Rules for selecting comparisons to be used in text generation (Milosavljevic)
• Principles for generating cases of particular difficulty levels (Carberry and Clarke)
• Procedure for matching requests for help with profiles of potential collaborators (Collins et al.)
• Rules for recommending forms of collaboration between students (Bull and Smith)

Interface techniques for communicating about the user model

• Techniques for making the student model inspectable and eliciting feedback on it (Bull)
• Techniques for presenting relevant parts of a user model to potential collaborators (Collins et al.)

4. Input for user model construction

Explicitly stated preferences, goals, etc.

• Critiques of proposed solutions (Gutkauf et al.; Paranagama et al.; Linden et al.)
• High-level visualization preferences (Bares and Lester)
• Interest in particular topics dealt with by documents (Benaki et al.)
• Preferences and goals concerning hypermedia presentations (Fink et al.)
• Explicit selection of hypertext contexts (Staff)
• Preferences and attitudes concerning aspects of language learning (Murphy and McTear)

Explicitly elicited information on personal characteristics

• Information about personal characteristics related to hypermedia use (Fink et al.)
• Job title, level of education, etc. (Collins et al.; Murphy and McTear; Strachan et al.)

Self-assessments

• Self-assessments of domain and system competence (Strachan et al.; De Carolis and Pizzutilo)
• Self-assessments of language proficiencies and motivation (Murphy and McTear)
• Self-reports on the successful completion of specific subtasks (Collins et al.)
• Self-reports on disabilities (Fink et al.)

Specific actions of the user

• History of dialog acts (Stein et al.)
• U's actions and locations within a large domain (Albrecht et al.)
• U's use of interface features and reading of hints about them (Brusilovsky and Schwarz)
• Hypermedia pages that U has visited (Weber and Specht; Milosavljevic; Staff; Akoulchina and Ganascia; Gori et al.; Maglio and Barrett)
• U's misspellings and ultimately chosen corrections (Spooner and Edwards)
• Aspects of disabled users' keyboard use (Trewin and Pain)
• Previous handling of instructional examples by U (Weber and Specht)
• Dialog actions in use of a hypermedia system (Fink et al.)
• Performance of particular tasks (Strachan et al.)
• Choices of device settings in various environments (Doux et al.)
• Aspects of behavior that reflect available working memory capacity (Schäfer and Weyrath)

Responses to test or practice items

• Responses to game-like ability tests (Gutkauf et al.)
• Answers to test questions handled individually or in collaboration (Bull and Smith)
• Answers to math problems; nature of hints required before answering (Beck et al.)
• Answers to test items in a tutoring system (Weber and Specht)
• Handling of previously presented medical cases (Carberry and Clarke)
• Performance on language test items (Murphy and McTear)
• Answers to test questions (Akoulchina and Ganascia)
• Answers to subtraction problems (Chiu et al.)
• Observable steps in a student's problem solving (Conati et al.)
• Performance of a student when given an opportunity to apply a given production rule (Corbett and Bhatnagar)
• Problem-solving actions within a training system (Moinard and Joab)
• Student actions or utterances that imply possession of a particular belief at a given time (Giangrandi and Tasso)
• Expressed interest in sample documents (Benaki et al.)

Other types of input

• Explicit assessments of U by a human instructor (Bull)
• U's medical record (Hirst et al.)
• Behavior with technical devices in various environments (Doux et al.)
• Student's performance on tests of declarative knowledge (Corbett and Bhatnagar)

5. Methods for constructing the user model

Bayesian methods

• Bayesian procedure for computing probabilities that production rules are known (Corbett and Bhatnagar)
• Bayesian networks for inferences about unobservable aspects of a student's problem solving (Conati et al.)
• Dynamic Bayesian networks for inferences about unobservable temporally variable properties (Schäfer and Weyrath)

Machine learning techniques

• Use of Input-Output Agent Modeling to derive a theory of a student's subtraction knowledge (Chiu et al.)
• Use of neural networks to adapt the system's profile of a decision maker (Paranagama et al.)
• Neural networks as an alternative technique for triggering stereotypes (Ambrosini et al.)
• Use of recurrent neural networks to summarize U's Web navigation behavior (Gori et al.)
• Use of a variant of the K-Means algorithm to classify users (Doux et al.)

Decision-theoretic techniques

• Principled method for elicitation and interpretation of critiques of proposed solutions (Linden et al.)

Stereotype-based techniques

• Ascription of properties associated with types of hypermedia users (Fink et al.)
• Ascription of WWW-related interests on the basis of user stereotypes (Ambrosini et al.)
• Derivation of initial proficiency estimates on the basis of U's overall level of advancement (Murphy and McTear)

Logic-based techniques

• Various inference techniques within a modal logic framework (Pohl and Höhle)
• Algorithms for making (nonmonotonic) inferences about beliefs held in particular time intervals (Giangrandi and Tasso)

Application-specific procedures for interpreting responses to test items

• Procedures for the interpretation of perceptual ability tests (Gutkauf et al.)
• Principle for inferring knowledge of concepts that are prerequisites for known concepts (Weber and Specht)
• Procedure for assessing U's domain expertise on the basis of U's answers to test questions (Akoulchina and Ganascia)
• Calculus for updating assessments of U's subskill proficiencies (Beck et al.)
• Computational procedures for estimating proficiencies and error-pronenesses (Murphy and McTear)
• Procedures for summarizing results of tests taken individually and in collaboration (Bull and Smith)
• Algorithm for generalizing the human instructor's assessments of U's strengths and weaknesses (Bull)
• Computation of declarative knowledge factor scores (Corbett and Bhatnagar)
• Comparison of a trainee's actions with those of an expert problem solving module (Moinard and Joab)

Other application-specific computations

• Updating of weights of incorrect spelling rules (Spooner and Edwards)
• Algorithm for updating assessments of system-related proficiency (Strachan et al.)

Application-specific qualitative rules

• Principles for inferring knowledge on the basis of dialog acts (Fink et al.)
• Rules for deriving low-level camera directives from visualization preferences (Bares and Lester)

6. Empirical foundations

Knowledge acquisition from domain experts

• Judgments of an expert surgeon concerning factors that influence the difficulty of medical cases (Carberry and Clarke)
• Retrospective thinking-aloud study of inferences by firemen about emergency callers (Schäfer and Weyrath)

Empirical studies conducted prior to system design

• Study of relationships between personality variables and decision making behavior (Paranagama et al.)
• Observation of Web navigation behavior (Maglio and Barrett)
• Experiments on relationships between expertise, presentation format, and comprehension by users (Kalyuga et al.)
• Derivation of conditional probability distributions for a Bayesian network from a database of observations (Albrecht et al.)
• Assessment of accuracy of knowledge tracing predictions that do not take declarative knowledge into account (Corbett and Bhatnagar)

Experience with real use of the system

• Responses to a flight recommendation system by Web users (Linden et al.)

Informal responses by early users

• Students' comments on an inspectable student model (Bull)
• Learners' responses to a commercial adaptive CALL system (Murphy and McTear)
• Users' responses to a conceptual-level programming environment (Seta et al.)

Empirical evaluations of systems

• Comparative evaluation of two systems' success in analyzing students' performance on subtraction problems (Chiu et al.)
• Comparison of the Recursive Modeling Method with simpler methods and with human performance (Noh and Gmytrasiewicz)
• Evaluation of a technique's performance on real and simulated data (Doux et al.)
• Assessment of accuracy of knowledge tracing predictions that take declarative knowledge into account (Corbett and Bhatnagar)
• Ratings of 3D visualizations produced on the basis of stated visualization preferences (Bares and Lester)
• Assessment of the appropriateness of keyboard adaptation recommendations (Trewin and Pain)
• Study of relationships among models of spelling behavior of different dyslexic writers (Spooner and Edwards)
• Evaluation of use of an adaptive chart-editing system (Gutkauf et al.)
• Study of effects of navigation support on students' motivation and the efficiency of their Web navigation (Weber and Specht)
• Formative evaluation of a math tutoring system (Beck et al.)
• Rating of adaptive and nonadaptive versions of a system by real users (Strachan et al.)
• Study of the feasability of the use of approximative inference algorithms (Conati et al.)
• Assessment by users of the relevance of documents supplied by an information filtering system (Ambrosini et al.)