Taking the Intentional Stance Seriously, or "Intending" to Improve Cognitive Systems

TL;DR

This paper discusses the importance of integrating human-like mental states, especially intentions, into cognitive systems to better align AI behavior with human expectations and improve user experience.

Contribution

It provides a detailed analysis of the mental state 'intention' and offers methodological suggestions for embedding such states into AI systems.

Findings

Analyzing 'intention' reveals key challenges in AI understanding.

Methodological approaches can facilitate the integration of mental states.

Aligning AI with folk-psychological theories improves human-AI interaction.

Abstract

Finding claims that researchers have made considerable progress in artificial intelligence over the last several decades is easy. However, our everyday interactions with cognitive systems (e.g., Siri, Alexa, DALL-E) quickly move from intriguing to frustrating. One cause of those frustrations rests in a mismatch between the expectations we have due to our inherent, folk-psychological theories and the real limitations we experience with existing computer programs. The software does not understand that people have goals, beliefs about how to achieve those goals, and intentions to act accordingly. One way to align cognitive systems with our expectations is to imbue them with mental states that mirror those we use to predict and explain human behavior. This paper discusses these concerns and illustrates the challenge of following this route by analyzing the mental state 'intention.' That analysis is joined with high-level methodological suggestions that support progress in this endeavor.