[Paper Review] A Measure Based Generalizable Approach to Understandability

Vikas Kushwaha, Sruti Srinivasa Ragavan+1 more|ArXiv.org|Mar 27, 2025

Complex Systems and Decision Making7 citations

TL;DR

Proposes a cognitive-science grounded, domain-agnostic framework of understandability with six dimensions to guide human–agent communication and steerability.

ABSTRACT

Successful agent-human partnerships require that any agent generated information is understandable to the human, and that the human can easily steer the agent towards a goal. Such effective communication requires the agent to develop a finer-level notion of what is understandable to the human. State-of-the-art agents, including LLMs, lack this detailed notion of understandability because they only capture average human sensibilities from the training data, and therefore afford limited steerability (e.g., requiring non-trivial prompt engineering). In this paper, instead of only relying on data, we argue for developing generalizable, domain-agnostic measures of understandability that can be used as directives for these agents. Existing research on understandability measures is fragmented, we survey various such efforts across domains, and lay a cognitive-science-rooted groundwork for more coherent and domain-agnostic research investigations in future.

Motivation & Objective

• Motivate the need for domain-agnostic understandability measures beyond data-driven priors.
• Organize understandability into six cognitive-dimension-based categories.
• Survey existing measures across domains and map them to the proposed dimensions.
• Lay groundwork for generalizable, implementable measures for agent guidance.

Proposed method

• Proposes six dimensions of understandability: perceptual advantage, memory cost, pattern decodability, cohesion, logical consistency, semantic fit.
• Reviews cognitive architectures and models of understanding to justify the dimensions.
• Connects domain-specific measures (e.g., text/code complexity, cohesion) to fundamental cognitive constructs.
• Outlines how these dimensions can be used as optimization directives for agents.
• Discusses potential for generalizable, domain-agnostic measures rooted in cognitive psychology.

Experimental results

Research questions

• RQ1 What fundamental cognitive dimensions underlie human understandability across artifacts (text, code, UI)?
• RQ2 How can domain-specific understandability measures be generalized using cognitive psychology principles?
• RQ3 What path forward exists for integrating these six dimensions into agent steering and optimization?

Key findings

• There exists a fragmented landscape of understandability measures across domains; a unified, six-dimension framework is proposed.
• The six dimensions map to stages of cognitive processing and can generalize across information artifacts.
• There is potential to use these dimensions as optimization directives for agents to improve steerability and comprehensibility.
• The framework provides a parsimonious, explanatory, and potentially predictive theory grounded in cognitive psychology.