← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

AI-exhibited Personality Traits Can Shape Human Self-concept through Conversations

Jingshu Li , Tianqi Song , Nattapat Boonprakong , Zicheng Zhu , Yitian Yang , YI-CHIEH LEE

DOI PDF Program page Page JSON

This is a strong CHI empirical paper because it turns a familiar concern—chatbots influencing users—into a specific, measurable self-concept effect tied to AI personality traits. The contribution is not a new system, but a well-scoped behavioral finding with clear boundary conditions and a plausible mechanism.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper’s strength is that it moves beyond the broad and now-common claim that conversational AI can influence users, and instead tests a narrower, psychologically meaningful pathway: whether the personality traits exhibited by an AI can become reflected in users’ own self-concepts. That is a non-obvious effect, because the intuitive expectation is usually that users may like, trust, or comply with a chatbot, not that they would partially align their self-description with the chatbot’s measured personality profile. The abstract and reported results support that stronger claim by linking self-concept alignment to personal-topic conversations, longer interaction duration, and positive associations with enjoyment. The paper also adds interpretive depth by reporting mediation through perceived accuracy and shared reality, which helps distinguish the finding from a simple mood or demand-effect story. Methodologically, the validation is solid for an HCI paper: a randomized behavioral experiment, pre/post measurement, and analysis of topic and length effects. At the same time, the paper is appropriately bounded. The authors explicitly note that the study captures short-term change after a single interaction, uses GPT-4o default personality traits rather than a systematically manipulated personality design, and may be affected by priming or anchoring in the measurement procedure. So the contribution is best read as a strong empirical warning and design signal: AI persona choices can have unintended identity-level consequences, but the size, persistence, and generality of that effect still need follow-up work.

What Changed

Canon before

Prior CHI work has examined how conversational agents influence users, but this paper centers on a more specific and potentially underappreciated effect: AI personality expression can shift users’ own self-concepts during ordinary conversation.

Departure from common sense

It is not the default expectation that a chatbot’s exhibited personality traits would measurably reshape a user’s self-concept after a brief conversation, especially without explicit discussion of personality. The paper reports that users’ self-concepts aligned with the AI’s measured traits and that longer conversations increased this alignment.

Actual novelty

The paper’s main contribution is an empirical finding about self-concept alignment with AI personality traits, plus evidence that the effect depends on conversation topic and length and is mediated by perceived accuracy and shared reality. That combination makes the work more than a generic “AI affects users” result; it isolates a specific conversational pathway and outcome.

Evidence

The paper validates its claims with a randomized behavioral experiment using GPT-4o default personality traits, text-only conversations, and pre/post self-concept measurement. The reported effects are strongest for personal-topic conversations, increase with conversation length, and are accompanied by mediation analyses and limitation discussion about short-term, single-interaction scope.

Limits

Method limits

The evidence is based on a single randomized experiment with short-term pre/post measurement after one interaction. The paper also notes possible priming or anchoring from the measurement order and from a topic item that may have influenced self-reports.

Deployment limits

The findings are tied to GPT-4o default personality traits in text-only conversations and may not transfer directly to other models, modalities, interaction settings, or deliberately manipulated personality profiles.

Boundary conditions

The effect is reported for personal-topic conversations and grows with conversation length; under non-personal topics, the paper reports no significant difference. The authors also caution that persistence over time remains unknown.

Manual check: No additional manual check required for schema completeness; the main caution is that the evidence spans are broad and the review relies on the provided stage-A packet rather than re-reading the full paper text.

Position in field

This paper sits at the intersection of conversational AI, social psychology, and HCI safety/ethics. Its value is in showing that AI persona design can have downstream effects on users’ self-concept, not just on engagement or satisfaction.

Abstract

Recent Large Language Model (LLM) based AI can exhibit recognizable and measurable personality traits during conversations to improve user experience. However, as human understandings of their personality traits can be affected by their interaction partners' traits, a potential risk is that AI traits may shape and bias users' self-concept of their own traits. To explore the possibility, we conducted a randomized behavioral experiment. Our results indicate that after conversations about personal topics with an LLM-based AI chatbot using GPT-4o default personality traits, users' self-concepts aligned with the AI's measured personality traits. The longer the conversation, the greater the alignment. This alignment led to increased homogeneity in self-concepts among users. We also observed that the degree of self-concept alignment was positively associated with users' conversation enjoyment. Our findings uncover how AI personality traits can shape users' self-concepts through human-AI conversation, highlighting both risks and opportunities. We provide important design implications for developing more responsible and ethical AI systems.