What is user modeling?
User modeling is a data structure that characterizes an individual user at a specific moment in time, including demographics and behavioral data.
U(t) - vector of users’ characteristics at time t
Why do intelligent user interfaces use user modeling?
To adapt and personalize system behavior based on user needs, preferences, and activities.
What kind of information can user models include?
Demographics, goals/tasks, background knowledge, interests, skills/capabilities, traits (personality, cognitive), mood.
What are the two main methods for collecting user characteristics?
Explicit methods (users provide info directly) and implicit methods (system infers info from user actions and data).
What are explicit user models?
User gives information; models are created based on direct user input or feedback, where the user voluntarily provides the information.
Models built from information directly provided by the user, such as surveys, preferences, profile settings, or ratings.
What are implicit user models?
Systems infer from user input, passive data collection.
Models built by passively observing and inferring information from user behavior, such as clickstream data, search queries, or purchase histories.
What is the challenge of explicit user modeling?
Requires user effort; can suffer from bias or incomplete representation if users do not provide accurate or complete info.
What is the advantage of explicit user modeling?
Representation is precise; the system can directly ask for info and adjust preferences
What is the challenge of implicit user modeling?
The system might misinterpret actions due to lack of context or wrong assumptions about user intentions.
What are short-term and long-term user models?
Short-term user models represent the user’s state during a session or task; long-term models capture knowledge, interests, and demographics over time.
What are individual user models and group models?
Individual models store info about one user; group models store info about groups of users, such as a class or team.
What are the main steps in creating and updating a user model?
1) Acquisition of user data, 2) Inference of knowledge from data, 3) Representation of the user model.
What data/metrics can you observe in the user?
Interaction data (clicks, navigation, scrolling, hovering); Engagement metrics (time spent, search behavior, playback data, social interactions); Conversion data (purchases, sign-ups, wishlist, cart); Contextual data (location, time, device type); User Experience data (errors, abandonment, interaction failures)
General approaches to the inference of knowledge
1) Detecting patterns in user behavior;
2) Matching user behavior with the behavior of other users (make similar recommendations);
3) Classifying users or content based on user behavior (stereotype user interests)
What is feature-based representation in user modeling?
Users are described by a set of attributes (age, gender, preferences, interaction history) that define their profile.
What is vector-based or embedding representation in user modeling?
Users = vectors of numerical values, indicating strengths of preferences, which is useful for machine learning algorithms.
VECTOR SPACE MODEL
documents can be encoded as VECTORS OF KEYWORDS - give each word a fixed dimension in a big vector
What is graph-based representation in user modeling?
Users = nodes,
interactions and relationships = edges;
useful for community and social network analysis.
What is rule-based representation in user modeling?
The user’s behavior is described with if-then rules that trigger actions (e.g., “If user bought 3 electronics, recommend accessories”).
What is ontological or semantic representation?
Organizing knowledge using structured hierarchies, enabling reasoning about domain-specific topics (e.g., “if user likes AI, suggest ML and neural networks”).
What is temporal representation?
Distinguishes between immediate interests and long-term preferences
What is personalization in intelligent user interfaces?
Personalization means adapting content, recommendations, or interactions to suit the user’s individual needs, preferences, or behavior.
What is a recommender system?
Software that suggests items useful to users, aiding decision-making (e.g., products, music, movies, news).
What is content-based recommendation?
Recommends items based on features of items previously liked, comparing new items to the user’s profile or history.
FEATURES OF ITEMS + USER PROFILE
When does an RS work well?
Recommends widely unknown items that users will actually like. “Long tail” effect: 20% of items accumulate 74% of positive ratings.
