---
title: "Chemical Analogies: Two Kinds of Explanation"
authors:
  - "Paul Thagard"
  - "Dawn M. Cohen"
  - "Keith J. Holyoak"
doi: ""
url: "https://ijcai.org/proceedings/1989/131"
pdf: "https://ijcai.org/proceedings/1989/0131.pdf"
venue: "Proceedings of the Eleventh International Joint Conference on Artificial Intelligence"
year: 1989
pages: "819-823"
track: "Machine Learning"
keywords:
---

# Chemical Analogies: Two Kinds of Explanation

Paul Thagard, Dawn M. Cohen, Keith J. Holyoak

*Proceedings of the Eleventh International Joint Conference on Artificial Intelligence*, pages 819-823.

## Abstract

Analogies are often used to help provide explanations of unfamiliar phenomena by comparing them to familiar phenomena. Analogical explanations are of two kinds: ones that provide systematic clarification, and ones that give a causal account of why something happened. We describe a theory and implementation of analogical mapping that applies to both kinds of explanation. The theory says that the elements of one analog are mapped onto the elements of another on the basis of structural, semantic, and pragmatic constraints. Our program ACME (Analogical Constraint Mapping Engine) uses localist networks of units representing mapping hypotheses to determine the correspondences between analogs. This paper describes ACME's application to eight analogies that have been used by chemistry teachers.