[ CogSci Summaries home | UP | email ]
http://www.jimdavies.org/summaries/

Kolodner, Janet L. and David Leake (1996). A Tutorial Introduction to Case-Based Reasoning . In Case-Based Reasoning: Experiences, Lessons and Future Directions, Leake, David (ed.). MIT Press. pp 31-65.

@InBook{,
  ALTauthor = 	 {},
  ALTeditor = 	 {},
  title = 	 {},
  chapter = 	 {},
  publisher = 	 {},
  year = 	 {},
  OPTkey = 	 {},
  OPTvolume = 	 {},
  OPTnumber = 	 {},
  OPTseries = 	 {},
  OPTtype = 	 {},
  OPTaddress = 	 {},
  OPTedition = 	 {},
  OPTmonth = 	 {},
  OPTpages = 	 {},
  OPTnote = 	 {},
  OPTannote = 	 {}
}

Author of the summary: Jake Auxier, 1999, jauxier1@cc.gatech.edu

Cite this paper for:

• Case-Based Reasoning: reasoning based on previous cases or experiences
• Problem solving case-based reasoning: the use of cases to propose solutions to problems
• Interpretive case-based reasoning: using cases for criticism and justification
• A case is a contextualized piece of knowledge representing an experience that teaches a lesson fundamental to achieving the goals of the reasoner
• The indexing problem is the problem of making sure that a case is accessed whenever appropriate

Introduction

Case-based reasoning advantages

• Allows the reasoner to propose solutions to problems quickly, avoiding the time necessary to derive those answers from scratch
• Allows a reasoner to propose solutions in domains that aren't completely understood
• Gives a reasoner a means of evaluating solutions when no algorithmic method is available for evaluation
• Cases make it possible to interpret open-ended and ill-defined concepts
• Cases help a reasoner to focus its reasoning on important parts of a problem by pointing out what features of a problem are the crucial ones
• Cases can warn of the potential for problems that have occurred in the past, alerting a reasoner to take actions to avoid repeating past mistakes

Case-based reasoning cycle:

1. Retrieve
2. Propose ballpark solution
3. Adaptation in problem solving CBR, Justification in Interpretive CBR
4. Criticism of the solution; may trigger further adaptation
5. Evaluation of the solution; may trigger further adaptation
6. Store the solution

General knowledge, when available, can guide the choice of indices, the determination of how well a new situation matches a stored case, and the choice of adaptation strategies.

Two major functional parts of a case: (1) the lesson(s) it teaches, (2) the context in which it can teach it lesson(s), described by its indexes; designating the circumstances in which it would be appropriately retrieved.

The Content of Cases and Indexing

 The content of a case is made up of: (1) the problem/situation description, (2) the solution, and (3) the outcome. The outcome is not needed but could be added to suggest solutions that work and use cases with failed solutions to warn of potential failures.

Indexes predict a case's usefulness. Indexes should be made that describe the tasks a case can be useful for. Indexes should be abstract enough to retrieve a relevant case in a variety of future situations, and indexes should be concrete enough to be easily recognizable in future situations.

Indexing Vocabularies

Indexing vocabularies have two parts: (1) a set of descriptive dimensions and (2) a set of values along each dimension.

Defining an Indexing Vocabulary

• The functional approach: look at what each case can be used for and the ways it needs to be described to make it available
• The reminding approach: examine the remindings that are natural among the human experts doing the designated task and look for similarities between new situations the experts are put in and the cases they are reminded of to find out which kinds of descriptors are the important ones for judging similarity and in what circumstances

Retrieval and Memory Update

Retrieval is done through matching and ranking procedures. Matching can use the indexes to find similar cases, but it needs to be able to distinguish which indexed features to focus on at any time.

Input to retrieval algorithms includes both a description of the new situation and also an indication of what the reasoner will use the case for.

Retrieval also depends on the retrieval algorithms used and situation assessment. Situation assessment is the process of analyzing a situation and elaborating it such that its description is in the same vocabulary as cases already in the case library.

Memory update is somewhat analogous to retrieval. Updating will look for a place to insert the new case rather than for a place to retrieve a similar case from.

Adaptation

Adaptation is when one takes a problem description and a solution that isn't quite right and manipulates the solution to make it better fit the problem description.

There are four general methods for adaptation:

1. Substitution: substitute values appropriate for the new situation for values in the old solution
2. Transformation: transform an old solution into one that will work in a new situation
3. Special-purpose adaptation and repair: used to carry out domain-specific and structure-modifying adaptations not covered by the above methods
4. Derivational replay: reuses the method for deriving an old solution or solution piece to derive a solution in anew situation.

Justification

This is used in interpretive CBR. Justification can be done by comparing new situations and prior cases, and by looking at what differentiates the new situation from old similar ones to ascertain whether or not the old interpretation is likely to hold.

Learning in Case-based Reasoning

A case-based reasoner's performance becomes more efficient by remembering old solutions and adapting them rather than having to derive answers from scratch each time.

Case-based reasoners become more competent over time, deriving better answers than thy could with less experience.

Where Learning Occurs

• Inductive formation of generalizations
• Through the accumulation of new cases
• Through the assignment and unassignment of indexes

Uses of CBR

Problem Solving CBR

• Design: CYCLOPS, KRITIK, KRITIK-2, JULIA, CADET, ARCHIE, ARCHIE-2, MIDAS, AskJef
• Planninng: CHEF, TRUCKER, EXPEDITOR, PLEXUS
• Diagnosis: CASEY, PROTOS

Interpretive CBR

• Classification: PROTOS
• Adversarial Reasoning: HYPO
• Projecting Effects: CSI Battle Planner, SCIED

Case-based Decision Aiding and Teaching

• ARCHI-2, SCIED, CSI Battle Planner, and AskJef

Types of Case-based System

Autonomous systems: solve problems by themselves; CHEF, JULIA, PLEXIUS, CASEY, PROTOS, HYPO.

Human-machine systems: work along with people to solve problems or interpret situations; CSI Battle Planner, ARCHIE-2, Clavier, SCIED.

Embedded systems: those in which a retrieval-only or autonomous case-based system is embedded as a component in a system with a larger purpose than problem solving or situation interpretation

Case Collection

• Cases should cover the range of reasoning tasks the system will be responsible for doing or supporting
• Over this range of reasoning tasks, cases should cover the range of well-known solutions and well-known mistakes
• Collecting cases is an incremental process

Summary author's notes:

• none

Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:

Last modified: 5/9/99 12:36:41 PM