mlsanityData to 7 Oct 2026

Metareasoning constraints couple narratives, affect and cognition

arXiv 2502.09487 ↗First posted 13 Feb 2025cs.CL cs.AI cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Narratives and emotions shape thoughts, and thoughts shape our feelings and stories we tell. Why narrative, affective and cognitive states interact remains unclear. We examine whether this mutual relationship reflects constraints on metareasoning - deciding what to think about - imposed by a shared computational state. Combining self-report and quantification of depression narratives using large language models, Study 1 (n=704) shows narrative state structure closely reflects the factorial structure in formal affect assessments, and that perturbation of the narrative state has commensurate effects on affect via a latent computational state. Study 2 (n=553) uses exposure to structured narratives to test model predictions causally in vivo. Narrative exposure has consistent effect on narrative states, with consequences on momentary mood, cognition, and affect. Critically, effects are predicted by latent computational state engagement. This supports the hypothesis that metareasoning constraints determine interactions between narratives, cognition and affect via a shared computational state.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 16 of its 118 references that are in the map. First posted on arXiv 13 Feb 2025.

Builds on

16 of 118 references.

References · 16

  1. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  2. 2023The Flan Collection: Designing Data and Methods for Effective Instruction Tuningcited by 734
  3. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 369
  4. 2022Data Distributional Properties Drive Emergent In-Context Learning in Transformerscited by 303
  5. 2023Role-Play with Large Language Modelscited by 290
  6. 2021Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factorscited by 106
  7. 2024HealMe: Harnessing Cognitive Reframing in Large Language Models for Psychotherapycited by 47
  8. 2023Theory of Mind in Large Language Models: Examining Performance of 11 State-of-the-Art models vs. Children Aged 7-10 on Advanced Testscited by 46
  9. 2021People construct simplified mental representations to plancited by 45
  10. 2023Understanding In-Context Learning via Supportive Pretraining Datacited by 41
  11. 2024The broader spectrum of in-context learningcited by 33
  12. 2024MindScape Study: Integrating LLM and Behavioral Sensing for Personalized AI-Driven Journaling Experiencescited by 24
4 more
  1. 2024LLM-based Conversational AI Therapist for Daily Functioning Screening and Psychotherapeutic Intervention via Everyday Smart Devicescited by 21
  2. 2022Using Natural Language and Program Abstractions to Instill Human Inductive Biases in Machinescited by 20
  3. 2024Rational Metareasoning for Large Language Modelscited by 20
  4. 2024Toward Large Language Models as a Therapeutic Tool: Comparing Prompting Techniques to Improve GPT-Delivered Problem-Solving Therapycited by 5

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

Similar, not cited

Close in wording, no citation either way

  1. 2025Large language models show fragile cognitive reasoning about human emotions
  2. 2023M-SENSE: Modeling Narrative Structure in Short Personal Narratives Using Protagonist's Mental Representations
  3. 2026Do Emotions in Prompts Matter? Effects of Emotional Framing on Large Language Models
  4. 2025Alignment, Exploration, and Novelty in Human-AI Interaction ↗
  5. 2026Affective Tools for Thought: Towards Shared Attention and Affective Reorienting in AI-Supported Thinking ↗
  6. 2026Directional Alignment and Narrative Agency in Human-LLM Co-Writing
  7. 2025Visually grounded emotion regulation via diffusion models and user-driven reappraisal
  8. 2025Beyond Context to Cognitive Appraisal: Emotion Reasoning as a Theory of Mind Benchmark for Large Language Models