Psychological Models of Explanatory Reasoning

This contribution contains a set of technical reports on psychological models of explanation, capturing different concepts, methods, and metrics.

Overview and Blogs

The XAI Literature Review
Non-Algorithmic Methods for XAI
Lessons for XAI from Intelligent Tutoring Systems
The AIQ Toolkit
The Discovery Platform
“Eggsplaining” AI: An Analogy for How Deep Nets Work

Actionable Concepts, Methods, and Metrics for Explanatory Reasoning

Validated scales for Explanation Goodness, Curiosity, and Trust

Metrics for Explainable AI
For explanation goodness, satisfaction and curiosity. Includes a review of methods for evaluating user mental models.
The Self-Explanation Scorecard
For evaluating the “explanatory depth” of explanations.
The Stakeholder Playbook
For tailoring explanations to stakeholder groups.
Measuring Trust in the XAI Context

Methodology for eliciting user mental models

The Mental Models Matrix

Methodology for creating “Cognitive Tutorials” for explaining AI systems

Cognitive Tutorial Authoring Guide

Guidance for experimental design for XAI evaluation

Methodology Requirements
Methodology Recommendations

Developing psychologically plausible models of explanation

Modeling the Explanation Process
The “Plausibility Gap” Model

A Computational Cognitive Model of explanatory reasoning

Computational Cognitive Model

A system for enabling users to share and discuss their own explanations

CXAI: Collaborative Explainable AI

A system for enabling developers to explore their systems: identify patterns, discover rules, find anomalies

The Discovery Platform

Updated: