Evaluating Developmental Cognition Capabilities of LLMs

2026-05-08

AI Summary

Current large language models struggle to adapt to users’ cognitive developmental stages, and existing evaluation methods are either difficult to scale or overly intrusive. Addressing this gap, this work proposes the first lightweight, open-source, self-administerable 20-item Developmental Sentence Completion Test (DSCT), grounded in Kegan’s Constructive Developmental Theory, to assess models’ capacity to recognize and generate text reflective of distinct developmental stages. Through sentence completion tasks, a developmental stage annotation framework, and multi-context prompting experiments, the study finds that state-of-the-art models accurately recover predefined developmental stages when simulating personas; achieve moderate agreement with real human responses, with adjacent-stage matches significantly outperforming exact matches; and, in unconditional generation, exhibit consistent intra-model-family developmental differences, with larger and more recent models tending to produce higher-order developmental content.

Abstract

Conversational AI is increasingly personalized around users' preferences, histories, goals, and knowledge, but much less around how users interpret and take up model outputs to construct and understand their reality. We draw on Robert Kegan's constructive-developmental theory as a complementary lens on this dimension. Existing methods for assessing developmental stage in the Keganian tradition rely either on expert interviews that do not scale or on sentence-completion instruments that are proprietary, lengthy, or invasive. To make this perspective tractable for LLM evaluation, we introduce the Developmental Sentence Completion Test (DSCT), a 20-item instrument designed to elicit developmental signal in self-administered text. Throughout, we treat the resulting labels as characterizations of stage-like structure in elicited responses, not as validated person-level developmental stage. We then ask how much of that signal can be recovered by LLMs across three elicited response regimes: simulated personas, real human respondents, and default model-generated answers. On simulated personas, top frontier models recover simulator-intended labels with high accuracy. On real human DSCT responses, human-LLM agreement is fair, with much stronger within-neighborhood than exact agreement. Finally, when LLMs answer DSCT prompts without persona-conditioning, their responses exhibit stable stage-like differences across model families, with larger and newer models tending to generate higher-rated text. These results suggest that stage-conditioned signal is cleaner in synthetic responses than in human-written DSCT text, and that the core constraint for stage-aware conversational AI is not classifier accuracy alone, but the availability of developmental signal from elicited text.

Problem

• developmental cognition
• large language models
• Kegan's theory
• sentence completion test
• conversational AI

Innovation

• developmental cognition
• large language models
• Kegan's theory
• sentence completion test
• stage-aware AI