My research program is grounded in the science of motivation, and my work investigates how motivation — from biological substrate to psychological experience — modulates learning and decision making. I am interested in how motivation can be regulated to support psychological flexibility and goal-directed behavior. This includes its volitional regulation, which can be learned with real-time fMRI neurofeedback from dopaminergic circuits.
Representative papers
Psychological flexibility as a measurable computational substrate for predicting successful self regulation and behavior change
Hakimi, in preparation
Temporal structure of learning to regulate ventral tegmental area using real-time neurofeedback
Hakimi et al., 2018 — Conference on Computational Cognitive Neuroscience
Enhanced neural responses to imagined primary rewards predict reduced monetary temporal discounting
Hakimi & Hare, 2015 — Journal of Neuroscience
Sustainable behavior change continues to be an elusive goal for many. I synthesize insights from multiple disciplines to develop new interventions that are not only more effective, but also more likely to work at for the right person, at the right time — at scale. This work spans multiple domains, including health, economic, and environmental decision making. This work is in dialogue with my work in applied settings, including the clinic and the workforce.
Representative papers
Unstuck: Ubiquitous just-in-time interventions to unleash creativity and sustain wellbeing of creative professionals
Paredes, Hakimi et al., 2026 — IMWUT
Machine learning reveals how personalized climate communication can both succeed and backfire
Harinen et al., 2022 — NeurIPS Workshop
Pairing facts with imagined consequences improves pandemic-related risk perception
Hakimi & Sinclair et al., 2021 — PNAS
Applying behavioral economics to improve adolescent and young adult health
Hakimi & Wong et al., 2020 — Journal of Adolescent Health
Creativity can feel like a nebulous process whereby an idea magically becomes a artifact. Demystifying the creative process with computational modeling can provide mechanistic insight has implications for both human creativity and human-AI co-creation. Precise models can also inform development of technologies that preserve and promote creative wellbeing over time.
Representative papers
The Moments-to-Meaning Framework: Designing technology for creative wellbeing across temporal scales
Hakimi & Kimani et al., in preparation
Designing rewards for rewarding designs: Demonstrating the impact of rewards on the creative design process
Nath et al., 2026 — DCC
Reassessing the value of creative activity traces for understanding and supporting creative work
Klenk et al., 2026 — CHI Workshop
Semantic properties of abstract prompts shape sequential decision making in design
Hakimi & Nandy et al., 2025 — RLDM
Semantic Properties of Word Prompts Shape Design Outcomes
Nandy et al., 2024 — DCC
One possible path to better human-AI alignment is generative AI systems with more structurally similar generative processes to humans, especially in time. This work applies multidisciplinary knowledge to the development of new systems that have the potential to be more useful to their human users.
Representative papers
Novel or surprising? Toward generating wowing designs with stable diffusion
Chong et al., 2026 — IDETC/CIE
Psychologically-inspired generative AI videos for supporting creativity
Hakimi et al., 2025 — ICCC
Generative AI for product esign: Getting the right design and the design right
Hong et al., 2023 — CHI Workshop
Why do people sometimes say one thing but do another? What do they really want? This work aims to improving measures of preference using psychology, physiology, and behavior, leveraging a combination of diverse sensors across time to better resolve individual preferences. These diverse, multimodal signals become inputs to new computational models that employ both machine learning and AI to improve the sensitivity and specificity of preference estimation and prediction. I aim for elicitation and modeling approaches that can both generalize and scale, thus providing foundational tools that support multiple lines of both basic and applied research.
Representative papers
Machine learning-based measure of cognitive complexity explains variance in rank-ordered preference
Hakimi et al., 2024 — CogSci
Understanding the cognitive complexity in language elicited by product images
Hakimi & Chen et al., 2024 — ICML Workshop
ConjointNet: Enhancing conjoint analysis for preference prediction with representation learning
Zhang et al., 2022 — IJCAI Workshop
Providing people support in the moment demands precise models of their internal state, including factors such as attention, motivation, and cognitive load. By inferring a person's state directly from behavior, future states and preferences can be simulated to improve human-AI teaming in shared tasks, such as driving.
Representative papers
Personalizing driver safety interfaces via driver cognitive factors inference
Sumner & DeCastro et al., 2024 — Scientific Reports
HMIway-env: A framework for simulating behaviors and preferences to support human-AI teaming in driving
Gopinath & DeCastro et al., 2022 — IEEE/CVF CVPRW
Learning latent traits for simulated cooperative driving tasks
DeCastro & Gopinath et al., 2022
Consumer demand is shaped by a variety of factors that are poorly captured by current data and modeling techniques. This work focuses on measuring the right things at the right time during the consumer journey in order to improve predictive models. Physiology — neural responses, in particular — can offer unique predictive power for forecasting. Combining neurophysiology with econometrics and machine learning affords new models that both increase forecast accuracy and reduce uncertainty.
Representative papers
Psychophysiological personas improve agentic prediction of consumer choice
Hakimi et al., in preparation
Brain activity forecasts changing market demand for innovative vehicles
Knutson et al., in preparation
AI-driven interpretable visual features for demand neuroforecasting
Eum et al., in preparation
How does the brain represent complex phenomena such as social decisions? Across domains, this research uses the neural substrates of such phenomena along with diverse complementary signals (e.g., genomics, pharmacology) to better predict individual outcomes.
Representative papers
Neural support for contributions of utility and narrative processing of evidence in juror decision making
Castrellon et al., 2022 — Journal of Neuroscience
Revealing the world of autism through the lens of a camera
Wang et al., 2016 — Current Biology
Vasopressin modulates social recognition-related activity in the left temporoparietal junction in humans
Zink et al., 2011 — Translational Psychiatry
Common alleles in OXTR impact prosocial temperament and human hypothalamic-limbic structure and function
Tost, Hakimi et al., 2010 — PNAS