Machine Learning in Design Taxonomy
- What can trigger learning?
- Failure; Success; Differences
- Need to improve abilities
- What are the elements supporting learning?
- critiques, praise, estimation, evaluation and advice
- sequences of design decisions
- traces of knowledge exchange and negotiation
- analysis of failures and conflicts
- feedback after completing the design task
- What might be learned?
- Constraints related to parameters or to elements of the design
- Dependencies between design parameters
- Support in favor of or against a decision
- Design rules
- Preferences in selection tasks
- Preconditions and postconditions for rules, actions and tasks
- Predictions about other agents
- Types of conflicts
- Heuristics to solve conflicts and to negotiate
- Availability of knowledge for learning
- Direct communication
- Indirect communication
- Central record of Design
- Repositories of design and interaction histories
- Methods of learning
- Explanation-based learning
- Induction
- Knowledge compilation
- Case-based learning
- Reinforcement learning
- Local vs. Global Learning
- Single designer
- Organization
- Consequences of learning
- Design improvement
- Improvement of the design process
Extracted from Grecu & Brown, Dimensions of Learning in Agent-based Design, AID'96, June 1996.