Seed Research Grants

Status

Closed

Date

Call for proposals will open in Summer 2025

Adversarial Examples for Humans

Automated Moderation of Small Group Deliberation

“Always On” Genetic Patient Diagnosis

Correcting Gender and Ethnic Biases in AI Algorithms

Dynamic Artificial Intelligence-Therapy for Autism on Google Glass

Enabling Natural-Language Interactions in Educational Software

Fast, Multiphase Human-in-the-loop Optimization of Exoskeleton Assistance

Free Exploration in Human-Centered AI Systems

Gender Bias in Conversations with Chatbots

Harnessing AI to Answer Questions about Diversity and Creativity

The Impact of Artificial Intelligence on Perceptions of Humanhood

The Impact on Society of Autonomous Mobile Robots: A Pilot Study

Improving Refugee Integration Through Data-Driven Algorithmic Assignment

Learning Behavior Change Interventions At Scale

Learning Decision Rules with Complex, Observational Data

Learning Haptic Feedback for Motion Guidance

Mining the Downstream Effects of Artificial Intelligence on How Clinicians Make Decisions

New Moral Economy in an Age of Artificial Intelligence

Novel Approach to Map Seasonal Changes in Infection Risk for Schistosomiasis: a Multi-Scale Integration of Satellite Data and Drone Imagery by Using Artificial Intelligence

Planning for Multi-Modal Human-Robot Communication

Scaling Collection of Labeled Data for Creating AI Systems through Observational Learning

Smart Learning Healthcare System for Human Behavior Change

Using AI to Facilitate Citizen Participation in Democratic Policy Deliberations

Using Computer Vision to Measure Neighborhood Variables Affecting Health

Using Deep Learning for Imaging Alzheimer’s Disease with Simultaneous Ultra-low-dose PET/MRI

Link copied to clipboard!

Stanford HAI Funds Groundbreaking AI Research Projects
Nikki Goth Itoi
Quick ReadJan 30
news
Thirty-two interdisciplinary teams will receive $2.37 million in Seed Research Grants to work toward initial results on ambitious proposals.
Policy-Shaped Prediction: Avoiding Distractions in Model-Based Reinforcement Learning
Nicholas Haber, Miles Huston, Isaac Kauvar
Dec 13
Research
Model-based reinforcement learning (MBRL) is a promising route to sampleefficient policy optimization. However, a known vulnerability of reconstructionbased MBRL consists of scenarios in which detailed aspects of the world are highly predictable, but irrelevant to learning a good policy. Such scenarios can lead the model to exhaust its capacity on meaningless content, at the cost of neglecting important environment dynamics. While existing approaches attempt to solve this problem, we highlight its continuing impact on leading MBRL methods —including DreamerV3 and DreamerPro — with a novel environment where background distractions are intricate, predictable, and useless for planning future actions. To address this challenge we develop a method for focusing the capacity of the world model through synergy of a pretrained segmentation model, a task-aware reconstruction loss, and adversarial learning. Our method outperforms a variety of other approaches designed to reduce the impact of distractors, and is an advance towards robust model-based reinforcement learning.
LABOR-LLM: Language-Based Occupational Representations with Large Language Models
Ayush Kanodia, Herman Brunborg, Keyon Vafa, Tianyu Du, Susan Athey
Dec 11
Research
Vafa et al. (2024) introduced a transformer-based econometric model, CAREER, that predicts a worker’s next job as a function of career history (an “occupation model”). CAREER was initially estimated (“pre-trained”) using a large, unrepresentative resume dataset, which served as a “foundation model,” and parameter estimation was continued (“fine-tuned”) using data from a representative survey. CAREER had better predictive performance than benchmarks. This paper considers an alternative where the resume-based foundation model is replaced by a large language model (LLM). We convert tabular data from the survey into text files that resemble resumes and fine-tune the LLMs using these text files with the objective to predict the next token (word). The resulting fine-tuned LLM is used as an input to an occupation model. Its predictive performance surpasses all prior models. We demonstrate the value of fine-tuning and further show that by adding more career data from a different population, fine-tuning smaller LLMs surpasses the performance of fine-tuning larger models.
How Persuasive Is AI-generated Propaganda?
Michael Tomz, Alex Stamos, Jason Chao, Josh A. Goldstein, Shelby Grossman
Feb 20
Research
Can large language models, a form of artificial intelligence (AI), generate persuasive propaganda? We conducted a preregistered survey experiment of US respondents to investigate the persuasiveness of news articles written by foreign propagandists compared to content generated by GPT-3 davinci (a large language model). We found that GPT-3 can create highly persuasive text as measured by participants’ agreement with propaganda theses. We further investigated whether a person fluent in English could improve propaganda persuasiveness. Editing the prompt fed to GPT-3 and/or curating GPT-3’s output made GPT-3 even more persuasive, and, under certain conditions, as persuasive as the original propaganda. Our findings suggest that propagandists could use AI to create convincing content with limited effort.
Sociotechnical Audits: Broadening the Algorithm Auditing Lens to Investigate Targeted Advertising
Danaë Metaxa, Michelle Lam, Ayush Pandit, Colin H. Kalicki, Poonam Sahoo, Rachit Gupta
Oct 04
Research
Algorithm audits are powerful tools for studying black-box systems without direct knowledge of their inner workings. While very effective in examining technical components, the method stops short of a sociotechnical frame, which would also consider users themselves as an integral and dynamic part of the system. Addressing this limitation, we propose the concept of sociotechnical auditing: auditing methods that evaluate algorithmic systems at the sociotechnical level, focusing on the interplay between algorithms and users as each impacts the other. Just as algorithm audits probe an algorithm with varied inputs and observe outputs, a sociotechnical audit (STA) additionally probes users, exposing them to different algorithmic behavior and measuring their resulting attitudes and behaviors. As an example of this method, we develop Intervenr, a platform for conducting browser-based, longitudinal sociotechnical audits with consenting, compensated participants. Intervenr investigates the algorithmic content users encounter online, and also coordinates systematic client-side interventions to understand how users change in response. As a case study, we deploy Intervenr in a two-week sociotechnical audit of online advertising (N = 244) to investigate the central premise that personalized ad targeting is more effective on users. In the first week, we observe and collect all browser ads delivered to users, and in the second, we deploy an ablation-style intervention that disrupts normal targeting by randomly pairing participants and swapping all their ads. We collect user-oriented metrics (self-reported ad interest and feeling of representation) and advertiser-oriented metrics (ad views, clicks, and recognition) throughout, along with a total of over 500,000 ads. Our STA finds that targeted ads indeed perform better with users, but also that users begin to acclimate to different ads in only a week, casting doubt on the primacy of personalized ad targeting given the impact of repeated exposure. In comparison with other evaluation methods that only study technical components, or only experiment on users, sociotechnical audits evaluate sociotechnical systems through the interplay of their technical and human components.
How Culture Shapes What People Want From AI
Hazel Markus, Daigo Misaki, Xiao Ge, Jeanne Tsai, Chunchen Xu
May 11
Research
There is an urgent need to incorporate the perspectives of culturally diverse groups into AI developments. We present a novel conceptual framework for research that aims to expand, reimagine, and reground mainstream visions of AI using independent and interdependent cultural models of the self and the environment. Two survey studies support this framework and provide preliminary evidence that people apply their cultural models when imagining their ideal AI. Compared with European American respondents, Chinese respondents viewed it as less important to control AI and more important to connect with AI, and were more likely to prefer AI with capacities to influence. Reflecting both cultural models, findings from African American respondents resembled both European American and Chinese respondents. We discuss study limitations and future directions and highlight the need to develop culturally responsive and relevant AI to serve a broader segment of the world population.
Minority-group incubators and majority-group reservoirs for promoting the diffusion of climate change and public health adaptations
Alyson L Singleton, Caroline Muraida, Cesar Augusto Lopez, Ian Harryman, James Holland Jones, Mallory J Harris, Matthew Adam Turner, Ronan Forde Arthur
Jan 01
Research
Current theory suggests that heterogeneous metapopulation structures can help foster the diffusion of innovations to solve pressing issues including climate change adaptation and promoting public health. In this paper, we develop an agent-based model of the spread of adaptations in simulated populations with minority-majority metapopulation structure, where subpopulations have different preferences for social interactions (i.e., homophily) and, consequently, learn deferentially from their own group. In our simulations, minority-majority-structured populations with moderate degrees of in-group preference better spread and maintained an adaptation compared to populations with more equal-sized groups and weak homophily. Minority groups act as incubators for novel adaptations, while majority groups act as reservoirs for the adaptation once it has spread widely. This suggests that population structure with in-group preference could promote the maintenance of novel adaptations.
Interaction of a Buoyant Plume with a Turbulent Canopy Mixing Layer
Hayoon Chung, Jeffrey R Koseff
Jun 23
Research
This study aims to understand the impact of instabilities and turbulence arising from canopy mixing layers on wind-driven wildfire spread. Using an experimental flume (water) setup with model vegetation canopy and thermally buoyant plumes, we study the influence of canopy-induced shear and turbulence on the behavior of buoyant plume trajectories. Using the length of the canopy upstream of the plume source to vary the strength of the canopy turbulence, we observed behaviors of the plume trajectory under varying turbulence yet constant cross-flow conditions. Results indicate that increasing canopy turbulence corresponds to increased strength of vertical oscillatory motion and variability in the plume trajectory/position. Furthermore, we find that the canopy coherent structures characterized at the plume source set the intensity and frequency at which the plume oscillates. These perturbations then move longitudinally along the length of the plume at the speed of the free stream velocity. However, the buoyancy developed by the plume can resist this impact of the canopy structures. Due to these competing effects, the oscillatory behavior of plumes in canopy systems is observed more significantly in systems where the canopy turbulence is dominant. These effects also have an influence on the mixing and entrainment of the plumes. We offer scaling analyses to find flow regimes in which canopy induced turbulence would be relevant in plume dynamics.

researchGrant

Seed Research Grants

Status

Closed

Date

Call for proposals will open in Summer 2025

Adversarial Examples for Humans

Gregory Valiant, Noah Goodman

Automated Moderation of Small Group Deliberation

Ashish Goel, James Fishkin

“Always On” Genetic Patient Diagnosis

Gill Bejerano, Jon Bernstein

Correcting Gender and Ethnic Biases in AI Algorithms

James Zou, Londa Schiebinger, Serena Yeung, Carlos Bustamante

Dynamic Artificial Intelligence-Therapy for Autism on Google Glass

Dennis Wall, Tom Robinson, Terry Winograd

Enabling Natural-Language Interactions in Educational Software

Alex Kolchinski, Sherry Ruan, Dan Schwartz, Emma Brunskill

Fast, Multiphase Human-in-the-loop Optimization of Exoskeleton Assistance

Steven Collins, Emma Brunskill

Free Exploration in Human-Centered AI Systems

Mohsen Bayati, Ramesh Johari

Gender Bias in Conversations with Chatbots

Katie Roehrick, Jeff Hancock, Byron Reeves, Londa Schiebinger, James Zou, Garrick Fernandez, Debnil Sur

Harnessing AI to Answer Questions about Diversity and Creativity

Dan McFarland, Londa Schiebinger, James Zou

The Impact of Artificial Intelligence on Perceptions of Humanhood

Benoît Monin, Erik Santoro

The Impact on Society of Autonomous Mobile Robots: A Pilot Study

Marco Pavone, Mark Duggan, David Grusky

Improving Refugee Integration Through Data-Driven Algorithmic Assignment

Jens Hainmueller, Kirk Bansak, Andrea Dillon, Jeremy Ferwerda, Dominik Hangartner, Duncan Lawrence, Jeremy Weinstein

Learning Behavior Change Interventions At Scale

Michael Bernstein, James Landay

Learning Decision Rules with Complex, Observational Data

Xinkun Nie, Stefan Wager

Learning Haptic Feedback for Motion Guidance

Julie Walker, Andrea Zanette, Mykel Kochenderfer, Allison Okamura

Mining the Downstream Effects of Artificial Intelligence on How Clinicians Make Decisions

Ron Li, Jason Ku Wang, Lance Downing, Lisa Shieh, Christopher Sharp, Jonathan Chen

New Moral Economy in an Age of Artificial Intelligence

Margaret Levi, Justice Mariano-Florentino Cuellar, Roberta Katz, John Markoff, Jane Shaw

Novel Approach to Map Seasonal Changes in Infection Risk for Schistosomiasis: a Multi-Scale Integration of Satellite Data and Drone Imagery by Using Artificial Intelligence

Giulio De Leo, Susanne Sokolow, Eric Lambin, Zac YC Liu, Chris Re, I. Jones, R. Grewelle, A. Ratner, A. Lund

Planning for Multi-Modal Human-Robot Communication

Yuhang Che, Cara Nunez, Allison Okamura, Dorsa Sadigh

Scaling Collection of Labeled Data for Creating AI Systems through Observational Learning

Daniel Rubin, Chris Re, Jared Dunnmon, Alex Ratner, Darvin Yi

Smart Learning Healthcare System for Human Behavior Change

Michelle Guo, Fei-Fei Li, Arnold Milstein

Using AI to Facilitate Citizen Participation in Democratic Policy Deliberations

Deger Turan, Frank Fukuyama, Jerry Kaplan, Larry Diamond, Eileen Donahoe, Chris Potts

Using Computer Vision to Measure Neighborhood Variables Affecting Health

Jackelyn Huang, Nikhil Naik

Using Deep Learning for Imaging Alzheimer’s Disease with Simultaneous Ultra-low-dose PET/MRI

Greg Zaharchuck, Bill Dally, John Pauly, Elizabeth Mormino

Link copied to clipboard!

Stanford HAI Funds Groundbreaking AI Research Projects
Nikki Goth Itoi
Quick ReadJan 30
news
Thirty-two interdisciplinary teams will receive $2.37 million in Seed Research Grants to work toward initial results on ambitious proposals.
Policy-Shaped Prediction: Avoiding Distractions in Model-Based Reinforcement Learning
Nicholas Haber, Miles Huston, Isaac Kauvar
Dec 13
Research
Model-based reinforcement learning (MBRL) is a promising route to sampleefficient policy optimization. However, a known vulnerability of reconstructionbased MBRL consists of scenarios in which detailed aspects of the world are highly predictable, but irrelevant to learning a good policy. Such scenarios can lead the model to exhaust its capacity on meaningless content, at the cost of neglecting important environment dynamics. While existing approaches attempt to solve this problem, we highlight its continuing impact on leading MBRL methods —including DreamerV3 and DreamerPro — with a novel environment where background distractions are intricate, predictable, and useless for planning future actions. To address this challenge we develop a method for focusing the capacity of the world model through synergy of a pretrained segmentation model, a task-aware reconstruction loss, and adversarial learning. Our method outperforms a variety of other approaches designed to reduce the impact of distractors, and is an advance towards robust model-based reinforcement learning.
LABOR-LLM: Language-Based Occupational Representations with Large Language Models
Ayush Kanodia, Herman Brunborg, Keyon Vafa, Tianyu Du, Susan Athey
Dec 11
Research
Vafa et al. (2024) introduced a transformer-based econometric model, CAREER, that predicts a worker’s next job as a function of career history (an “occupation model”). CAREER was initially estimated (“pre-trained”) using a large, unrepresentative resume dataset, which served as a “foundation model,” and parameter estimation was continued (“fine-tuned”) using data from a representative survey. CAREER had better predictive performance than benchmarks. This paper considers an alternative where the resume-based foundation model is replaced by a large language model (LLM). We convert tabular data from the survey into text files that resemble resumes and fine-tune the LLMs using these text files with the objective to predict the next token (word). The resulting fine-tuned LLM is used as an input to an occupation model. Its predictive performance surpasses all prior models. We demonstrate the value of fine-tuning and further show that by adding more career data from a different population, fine-tuning smaller LLMs surpasses the performance of fine-tuning larger models.
How Persuasive Is AI-generated Propaganda?
Michael Tomz, Alex Stamos, Jason Chao, Josh A. Goldstein, Shelby Grossman
Feb 20
Research
Can large language models, a form of artificial intelligence (AI), generate persuasive propaganda? We conducted a preregistered survey experiment of US respondents to investigate the persuasiveness of news articles written by foreign propagandists compared to content generated by GPT-3 davinci (a large language model). We found that GPT-3 can create highly persuasive text as measured by participants’ agreement with propaganda theses. We further investigated whether a person fluent in English could improve propaganda persuasiveness. Editing the prompt fed to GPT-3 and/or curating GPT-3’s output made GPT-3 even more persuasive, and, under certain conditions, as persuasive as the original propaganda. Our findings suggest that propagandists could use AI to create convincing content with limited effort.
Sociotechnical Audits: Broadening the Algorithm Auditing Lens to Investigate Targeted Advertising
Danaë Metaxa, Michelle Lam, Ayush Pandit, Colin H. Kalicki, Poonam Sahoo, Rachit Gupta
Oct 04
Research
Algorithm audits are powerful tools for studying black-box systems without direct knowledge of their inner workings. While very effective in examining technical components, the method stops short of a sociotechnical frame, which would also consider users themselves as an integral and dynamic part of the system. Addressing this limitation, we propose the concept of sociotechnical auditing: auditing methods that evaluate algorithmic systems at the sociotechnical level, focusing on the interplay between algorithms and users as each impacts the other. Just as algorithm audits probe an algorithm with varied inputs and observe outputs, a sociotechnical audit (STA) additionally probes users, exposing them to different algorithmic behavior and measuring their resulting attitudes and behaviors. As an example of this method, we develop Intervenr, a platform for conducting browser-based, longitudinal sociotechnical audits with consenting, compensated participants. Intervenr investigates the algorithmic content users encounter online, and also coordinates systematic client-side interventions to understand how users change in response. As a case study, we deploy Intervenr in a two-week sociotechnical audit of online advertising (N = 244) to investigate the central premise that personalized ad targeting is more effective on users. In the first week, we observe and collect all browser ads delivered to users, and in the second, we deploy an ablation-style intervention that disrupts normal targeting by randomly pairing participants and swapping all their ads. We collect user-oriented metrics (self-reported ad interest and feeling of representation) and advertiser-oriented metrics (ad views, clicks, and recognition) throughout, along with a total of over 500,000 ads. Our STA finds that targeted ads indeed perform better with users, but also that users begin to acclimate to different ads in only a week, casting doubt on the primacy of personalized ad targeting given the impact of repeated exposure. In comparison with other evaluation methods that only study technical components, or only experiment on users, sociotechnical audits evaluate sociotechnical systems through the interplay of their technical and human components.
How Culture Shapes What People Want From AI
Hazel Markus, Daigo Misaki, Xiao Ge, Jeanne Tsai, Chunchen Xu
May 11
Research
There is an urgent need to incorporate the perspectives of culturally diverse groups into AI developments. We present a novel conceptual framework for research that aims to expand, reimagine, and reground mainstream visions of AI using independent and interdependent cultural models of the self and the environment. Two survey studies support this framework and provide preliminary evidence that people apply their cultural models when imagining their ideal AI. Compared with European American respondents, Chinese respondents viewed it as less important to control AI and more important to connect with AI, and were more likely to prefer AI with capacities to influence. Reflecting both cultural models, findings from African American respondents resembled both European American and Chinese respondents. We discuss study limitations and future directions and highlight the need to develop culturally responsive and relevant AI to serve a broader segment of the world population.
Minority-group incubators and majority-group reservoirs for promoting the diffusion of climate change and public health adaptations
Alyson L Singleton, Caroline Muraida, Cesar Augusto Lopez, Ian Harryman, James Holland Jones, Mallory J Harris, Matthew Adam Turner, Ronan Forde Arthur
Jan 01
Research
Current theory suggests that heterogeneous metapopulation structures can help foster the diffusion of innovations to solve pressing issues including climate change adaptation and promoting public health. In this paper, we develop an agent-based model of the spread of adaptations in simulated populations with minority-majority metapopulation structure, where subpopulations have different preferences for social interactions (i.e., homophily) and, consequently, learn deferentially from their own group. In our simulations, minority-majority-structured populations with moderate degrees of in-group preference better spread and maintained an adaptation compared to populations with more equal-sized groups and weak homophily. Minority groups act as incubators for novel adaptations, while majority groups act as reservoirs for the adaptation once it has spread widely. This suggests that population structure with in-group preference could promote the maintenance of novel adaptations.
Interaction of a Buoyant Plume with a Turbulent Canopy Mixing Layer
Hayoon Chung, Jeffrey R Koseff
Jun 23
Research
This study aims to understand the impact of instabilities and turbulence arising from canopy mixing layers on wind-driven wildfire spread. Using an experimental flume (water) setup with model vegetation canopy and thermally buoyant plumes, we study the influence of canopy-induced shear and turbulence on the behavior of buoyant plume trajectories. Using the length of the canopy upstream of the plume source to vary the strength of the canopy turbulence, we observed behaviors of the plume trajectory under varying turbulence yet constant cross-flow conditions. Results indicate that increasing canopy turbulence corresponds to increased strength of vertical oscillatory motion and variability in the plume trajectory/position. Furthermore, we find that the canopy coherent structures characterized at the plume source set the intensity and frequency at which the plume oscillates. These perturbations then move longitudinally along the length of the plume at the speed of the free stream velocity. However, the buoyancy developed by the plume can resist this impact of the canopy structures. Due to these competing effects, the oscillatory behavior of plumes in canopy systems is observed more significantly in systems where the canopy turbulence is dominant. These effects also have an influence on the mixing and entrainment of the plumes. We offer scaling analyses to find flow regimes in which canopy induced turbulence would be relevant in plume dynamics.