Recalled Emotion Regulation in Matched Human and LLM Chatbot Support Episodes
Abstract
People increasingly turn to LLM chatbots for emotional support, but little research compares that support with human support in the same framework. We distinguish cognitive-symbolic support (articulation, validation, reframing), which a language model can deliver, from embodied-relational support (reciprocal presence, co-regulation, being understood by a real other), which it cannot, and predicted similar outcomes on the first and a human advantage on the second. In a preregistered within-subject retrospective study (osf.io/pesrh), 583 Prolific participants who had talked to both a person and an LLM chatbot about the same or a comparable personal topic in the past week reported both episodes in randomized order. Outcomes were recalled distress change, perceived attunement, emotional expression and processing, and help-seeking. Both episode types were associated with large recalled distress reductions (AI dz = 1.22; human dz = 1.00), and the difference between them met the preregistered equivalence criterion (dz = 0.20; pTOST = .009). Contrary to prediction, attunement, expression, processing, and same-source help-seeking intention showed no detectable difference by source, with 95 percent confidence intervals within dz = ±0.12. Conclusions held when the 479 excluded cases were restored and when topic domain was covaried. Felt closeness to the source, higher for humans, predicted attunement in both episode types, processing in human episodes, and distress reduction in AI episodes. A reflexive thematic analysis of 619 topic descriptions identified four needs that participants brought to both sources. Among recent dual-source users, recalled support outcomes were strikingly similar for a person and a chatbot. Whether the underlying relational processes or longer-term outcomes are similar remains untested.
Related
- Imposter Syndrome Among Students of Color: The Role of Perfectionism and Emotion Regulation
- Associations of sleep disturbance and emotion regulation with perceived workplace safety and fatigue among male firefighters with prior occupational accidents : A cross-sectional study
- The Role of Parental Reflective Functioning in the Association Between Parental and Youth Emotion Regulation Among Youth With Serious Mental Health Challenges
- Everyday use of AI for emotion regulation depends on the person and the situation
- Associations Among family environment, self-identity, cognitive emotion regulation strategy use, and physical activity in adolescents: A multi-theory integration perspective
BibTeX
@misc{maruani2026recalled,
title = {{Recalled Emotion Regulation in Matched Human and LLM Chatbot Support Episodes}},
author = {Luigi Maruani and June Thompson and Maahir Uttam},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/y65d4_v2},
url = {https://osf.io/preprints/psyarxiv/y65d4_v2/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Recalled Emotion Regulation in Matched Human and LLM Chatbot Support Episodes
AU - Luigi Maruani
AU - June Thompson
AU - Maahir Uttam
PY - 2026
DA - 2026/09/09
PB - PsyArXiv
DO - 10.31234/osf.io/y65d4_v2
UR - https://osf.io/preprints/psyarxiv/y65d4_v2/
AB - People increasingly turn to LLM chatbots for emotional support, but little research compares that support with human support in the same framework. We distinguish cognitive-symbolic support (articulation, validation, reframing), which a language model can deliver, from embodied-relational support (reciprocal presence, co-regulation, being understood by a real other), which it cannot, and predicted similar outcomes on the first and a human advantage on the second. In a preregistered within-subject retrospective study (osf.io/pesrh), 583 Prolific participants who had talked to both a person and an LLM chatbot about the same or a comparable personal topic in the past week reported both episodes in randomized order. Outcomes were recalled distress change, perceived attunement, emotional expression and processing, and help-seeking. Both episode types were associated with large recalled distress reductions (AI dz = 1.22; human dz = 1.00), and the difference between them met the preregistered equivalence criterion (dz = 0.20; pTOST = .009). Contrary to prediction, attunement, expression, processing, and same-source help-seeking intention showed no detectable difference by source, with 95 percent confidence intervals within dz = ±0.12. Conclusions held when the 479 excluded cases were restored and when topic domain was covaried. Felt closeness to the source, higher for humans, predicted attunement in both episode types, processing in human episodes, and distress reduction in AI episodes. A reflexive thematic analysis of 619 topic descriptions identified four needs that participants brought to both sources. Among recent dual-source users, recalled support outcomes were strikingly similar for a person and a chatbot. Whether the underlying relational processes or longer-term outcomes are similar remains untested.
ER -