Lecture 4 - Problem Solving Flashcards
What is a problem?
What is a problem?
Start state: your current situation
Goal state: the desired situation
It is not clear how to get from the start to the goal – subjective
What is a Well Defined Problem?
All aspects of the problem are well defined
Initial state, goal state and possible moves
What is an ill defined problem?
Start and end state or the possible strategies may be unknown
Most everyday problems
What is a knowledge rich problem?
Problems that require specific knowledge
What is a knowledge lean problem?
Problems that do not require specific knowledge
Puzzles
The Behaviorist approach
Trial and error learning – Thorndike’s (1898) cat experiment • Unsystematic behaviour • Requires no knowledge • Slow • Doesn’t work for all problems • Risky
Gestalt Approach
Problem solving requires insight
An aha moment
Evaluation of Gestalt Approach
Recognises the role of insight
Mechanisms underlying insight are not specified
Representational Change Theory
Aims to explain the processes underlying insight
- Construct a problem representation
- Retrieve operators (moves/actions) from memory by spreading activation from the problem representation
Limitations of the Representational Change Theory
Does not explain what leads to representational change
Why incubation help
Info processing - Computational modelling approach – general problem solver
Problem solving is a search though the problem space
Don’t have working memory capacity to think of all the possible moves
Objective measure of optimal performance and can test whether people make moves that are consistent with the heuristics
Info Processing Heuristics – Newell & Simon, 1972
Hill climbing – choose a move that brings you closer to the goal
This can lead to problems if the next step does not bring us closer to the goal – intermediate peaks or plateaus
Means end analysis - make achievable subgoals
Info Processing Evaluation
Works well for well defined problems Good objective measures of how well people perform Led to well-specified computer models Many everyday problem ill defined Doesn’t work for insight problems
Analogical Problem Solving - learning from past problems
Negative transfer – functional fixedness
Positive transfer – near transfer to similar context, far transfer to different context
Learning from analogy – surface and structural features
Expertise
Focus on problems that depend on expertise knowledge and learning
Chess expertise – compare experts and masters
