OUR POSITION
AI systems already respond to psychological signals in user language. Making those signals explicit makes AI systems more effective, safer, and more aligned with the people using them.
Every AI system forms inferences about user state between input and output. Those inferences influence every response — and are invisible to every evaluation pipeline, safety system, and fine-tuning process in use today.
Our research starts from a single premise: when user psychological state becomes an explicit, structured variable rather than an opaque internal inference, AI systems become more effective, safer, and more aligned with the people using them. We are building the evidence base across domains where that premise can be tested, measured, and published.
ACTIVE RESEARCH
Receptiviti conducts research and experiments that contribute to the questions AI safety, alignment, and evaluation teams are actively working on.
SCOPED
Interaction state as a missing dimension in AI evaluation
Standard AI evaluation measures what the model produces — accuracy, helpfulness, harmlessness. It does not measure what the interaction does to the person.
AI systems condition their responses on psychological signals inferred from the user's language. That inference happens between input and output and is invisible to standard eval pipelines, which see what went in and what came out — but not what the model understood about the person in between. Understanding how AI understands users may be one of the more tractable paths to understanding AI itself.
This study makes the case for human-state signals as a first-class eval criterion alongside accuracy, helpfulness, and harmlessness — and proposes the measurement framework for doing so.
DOMAIN
AI evaluation & interpretability
PARTNERS
Seeking AI teams to co-validate
SCOPED
Providing measured user state as context to educational AI
A GPT Study Mode experiment conditioning AI tutor behaviour on explicit interaction-state variables — cognitive load, confidence, analytical engagement — produced a 4% improvement in overall educational effectiveness across 25 blinded evaluations. Largest gains in reasoning and scaffolding (+6.3%) and cognitive load management (+5.9%).
This study is designed to validate that finding at scale, across subject domains and learner profiles, and to characterise which psychological dimensions drive the largest gains.
DOMAIN
Educational AI
PARTNERS
Education platforms and AI tutoring teams
SCOPED
Providing measured user state as context to customer service AI
Customer support AI is optimised for resolution rates, handle time, and CSAT. None of these metrics capture what is happening to the customer during the interaction — whether frustration is building, confidence is eroding, or rapport is being established or lost.
This study asks whether AI agents conditioned on real-time interaction state deliver better resolution outcomes and lower escalation rates than agents operating on inference alone.
DOMAIN
Enterprise AI
PARTNERS
Enterprise AI deployment teams
PUBLISHED
Selected published research by Dr. Molly Ireland, Head of Social Psychology at Receptiviti, and by Ryan Boyd, Assistant Professor of Psychology at UT Dallas and academic advisor to Receptiviti.
PUBLISHED: PNAS Nexus, 2024
Large language models display human-like social desirability biases in personality surveys
Salecha, Ireland et al.,2024. Large language models display human-like response biases when they infer they are being evaluated, with effects up to 1.20 human SD across GPT-4, Claude 3, Llama 3, and PaLM-2. Co-authored by Molly Ireland (Receptiviti).
PUBLISHED: npj Artificial Intelligence · 2026
PsychAdapter: adapting LLMs to reflect traits, personality, and mental health
Vu, Boyd, Eichstaedt et al., 2026. A lightweight LLM architectural modification that generates text reliably reflecting Big Five personality traits (87.3% accuracy) and mental health variables (96.7% accuracy). Co-authored by Ryan Boyd (UT Dallas), academic partner to Receptiviti.
PUBLISHED: npj Mental Health Research, 2025
Psychosocial dynamics of suicidality and nonsuicidal self-injury: a digital linguistic perspective
Entwistle, Hoemann, Nightingale & Boyd, 2025. Large-scale naturalistic study of the language dynamics surrounding suicidality and self-injury in 992 individuals with borderline personality disorder (66,786 posts). Co-authored by Ryan Boyd (UT Dallas), academic advisor to Receptiviti.
PUBLISHED: Perspectives on Psychological Science, 2026
Artificial intelligence and the psychology of human connection
Boyd & Markowitz, 2026. Introduces the MIRA model - a theoretical framework for when and how AI functions as a relational entity in human ecosystems. Language is the primary modality through which that relationship operates. Co-authored by Ryan Boyd (UT Dallas), academic advisor to Receptiviti.
PUBLISHED: Clinical Psychological Science, 2026
Replicability and validity of a new AI assessment of PTSD from patient language
Kjell, Ganesan, Boyd et al., 2026. AI-based psychological assessment from language, when grounded in validated measurement, produces replicable, clinically meaningful results across sequential evaluation with preregistered models.
Co-authored by Ryan Boyd (UT Dallas), academic advisor to Receptiviti.
PREPRINT: arXiv, May 2026, under review
When support escalates distress: regulation and escalation in LLM responses to venting and advice-seeking
Chi, Ganesan, Boyd, Ungar & Guntuku, 2026. Across 178,800 Reddit posts, LLM responses to venting simultaneously regulate and escalate distress — and the escalation is invisible to standard safety evaluations. Therapist personas reduce escalation; friend personas increase both. Measured using LIWC-22. Co-authored by Ryan Boyd (UT Dallas), academic advisor to Receptiviti.