HAI and Accelerator for Learning Partnership Grant | Stanford HAI
Stanford
University
  • Stanford Home
  • Maps & Directions
  • Search Stanford
  • Emergency Info
  • Terms of Use
  • Privacy
  • Copyright
  • Trademarks
  • Non-Discrimination
  • Accessibility
© Stanford University.  Stanford, California 94305.
Skip to content
  • About

    • About
    • People
    • Get Involved with HAI
    • Support HAI
    • Subscribe to Email
  • Research

    • Research
    • Fellowship Programs
    • Grants
    • Student Affinity Groups
    • Centers & Labs
    • Research Publications
    • Research Partners
  • Education

    • Education
    • Executive and Professional Education
    • Government and Policymakers
    • K-12
    • Stanford Students
  • Policy

    • Policy
    • Policy Publications
    • Policymaker Education
    • Student Opportunities
  • AI Index

    • AI Index
    • AI Index Report
    • Global Vibrancy Tool
    • People
  • News
  • Events
  • Industry
  • Centers & Labs
Navigate
  • About
  • Events
  • Careers
  • Search
Participate
  • Get Involved
  • Support HAI
  • Contact Us

Stay Up To Date

Get the latest news, advances in research, policy work, and education program updates from HAI in your inbox weekly.

Sign Up For Latest News

researchGrant

HAI and Accelerator for Learning Partnership Grant

Status
Closed
Apply
Grant Overview
2024 Grant Recipients
2023 Generative AI for the Future of Learning Recipients

Generative AI for the Future of Learning

The Stanford Institute for Human-Centered Artificial Intelligence and the Stanford Accelerator for Learning invited proposals for designs and/or studies that explore how generative AI can be applied in novel ways to support learning or investigate critical issues in learning contexts.

Faculty, academic staff, postdocs, and students from across the university were invited to apply for seed grant awards ranging from $5,000 to $100,000.

Generative AI, or artificial intelligence that can generate novel content through text, audio files, or images, has reached a tipping point where it can produce high-quality output that can support many different kinds of tasks. For example, ChatGPT can write essays and code, DALL-E can create images and art, while other forms of generative AI can produce recipes, music, and videos. These new forms of generative AI have the capacity to change how we think, create, teach, and also learn. They may also change our perspective on what is important to learn.

In the future, we expect Stanford will be a hub for generative AI-based research and solutions that are scalable. This seed grant will fund early and exploratory stages of this work, such as designs, prototypes, and pilot studies that may have the potential to scale or have broad impact in the future. Proposals may target any type of learner (e.g., worker, student, teacher, family) in any setting (e.g., workplaces, museums, classrooms, homes).

Request for proposals: Closed

Grant Overview
2024 Grant Recipients
2023 Generative AI for the Future of Learning Recipients
Share
Link copied to clipboard!
Related
  • A Large Scale RCT on Effective Error Messages in CS1
    Sierra Wang, John Mitchell, Christopher Piech
    Mar 07
    Research

    In this paper, we evaluate the most effective error message types through a large-scale randomized controlled trial conducted in an open-access, online introductory computer science course with 8,762 students from 146 countries. We assess existing error message enhancement strategies, as well as two novel approaches of our own: (1) generating error messages using OpenAI's GPT in real time and (2) constructing error messages that incorporate the course discussion forum. By examining students' direct responses to error messages, and their behavior throughout the course, we quantitatively evaluate the immediate and longer term efficacy of different error message types. We find that students using GPT generated error messages repeat an error 23.1% less often in the subsequent attempt, and resolve an error in 34.8% fewer additional attempts, compared to students using standard error messages. We also perform an analysis across various demographics to understand any disparities in the impact of different error message types. Our results find no significant difference in the effectiveness of GPT generated error messages for students from varying socioeconomic and demographic backgrounds. Our findings underscore GPT generated error messages as the most helpful error message type, especially as a universally effective intervention across demographics.

  • Evaluating Human and Machine Understanding of Data Visualizations
    Arnav Verma, Kushin Mukherjee, Christopher Potts, Elisa Kreiss, Judith Fan
    Jan 01
    Research
    Your browser does not support the video tag.

    Although data visualizations are a relatively recent invention, most people are expected to know how to read them. How do current machine learning systems compare with people when performing tasks involving data visualizations? Prior work evaluating machine data visualization understanding has relied upon weak benchmarks that do not resemble the tests used to assess these abilities in humans. We evaluated several state-of-the-art algorithms on data visualization literacy assessments designed for humans, and compared their responses to multiple cohorts of human participants with varying levels of experience with high school-level math. We found that these models systematically underperform all human cohorts and are highly sensitive to small changes in how they are prompted. Among the models we tested, GPT-4V most closely approximates human error patterns, but gaps remain between all models and humans. Our findings highlight the need for stronger benchmarks for data visualization understanding to advance artificial systems towards human-like reasoning about data visualizations.

Educational simulations such as PhET can increase learner engagement and promote generative learning. Yet, it is nearly impossible for educators to create their own interactive simulations to support their learners. These scholars will develop the Simulations Adaptive Learning Tool (SALT), a tool that leverages generative AI capabilities to allow educators to create or adapt interactive content to meet learners' needs. Using SALT, educators can personalize the simulations in ways that nurture students' cultural and linguistic diversity, which can enhance the effectiveness of their learning experience.

Name

Role

School

Department

Hari Subramonyam

Main PI

Graduate School of Education

Graduate School of Education

Nick Haber

Co-PI

Graduate School of Education

Graduate School of Education

Shima Salehi

Co-PI

Graduate School of Education

Graduate School of Education

Maneesh Agrawala

Co-PI

School of Engineering

Computer Science

Roy Pea

Co-PI

Graduate School of Education

Graduate School of Education

Data visualizations are indispensable for communicating patterns in quantitative data and are crucial in STEM learning contexts. Unfortunately, these visualizations are only rarely accompanied by high-quality descriptions that would make them more accessible to blind and low-vision learners. These scholars will use research in AI, cognition, and education to make complex data visualizations more accessible through: development of datasets containing high-quality descriptions of many kinds of data visualizations; training of AI systems that generate descriptions for novel data visualizations; and measurement of the impact of human and model-generated descriptions on learner comprehension.

Name

Role

School

Department

Christopher Potts

Main PI

School of Humanities and Sciences

Linguistics

Judith Fan

Co-PI

School of Humanities and Sciences

Psychology

Providing timely, personalized and mindset-supportive feedback to students is an integral part of high quality instruction, yet it is a nontrivial and extremely time intensive task. These scholars will develop Teach M-Powered, a generative AI-powered tool that assists teachers with writing effective feedback to students.

Name

Role

School

Department

Dora Demszky

Main PI

Graduate School of Education

Graduate School of Education

This project aims to explore various approaches to AI-assisted collaborative learning and develop and evaluate sample uses over the coming year. The team will develop an action research community where students collaborate on toolkits that will be offered to Stanford project partners for their teaching and learning environments next academic year, while examining gen-AI ethical principles and AI challenges for education.

Name

Role

School

Department

John Mitchell

Main PI

School of Engineering

Computer Science

Jennifer Langer-Osuna

Co-PI

Graduate School of Education

Graduate School of Education

Since ChatGPT was released in November 2022 and many other models followed, researchers have studied their inability to generate African American English (AAE) in conversation with Black student communities. Such a deficit arises from the corpora that commercial generative models deploy. These scholars will build on the TwitterAAE and CORAAL corpora with their own data set called BlackRhetorics and use NLP transfer learning and dialect techniques to improve the tools for Black student research. This Black research team demonstrates how to deploy generative models for inclusive Black language pedagogies.

Name

Role

School

Department

Adam Banks

Main PI

Graduate School of Education

Graduate School of Education

Harriet Jernigan

Co-PI

Vice Provost for Undergraduate Education

Writing and Rhetoric

This project addresses the challenge of limited access to quality education and software in the rapidly growing field of biomedical data, which generates vast amounts of data requiring advanced computational skills to process. The team proposes expanding the Stanford Data Ocean (SDO) platform (shorturl.at/aIJKS) with AI Chatbots like ChatGPT to support interdisciplinary concepts in learning Precision Medicine. Their integrated curricula will be customized to address major challenges in accessing quality education for underserved communities.

Name

Role

School

Department

Michael Snyder

Main PI

School of Medicine

Genetics

Anshul Kundaje

Co-PI

School of Medicine

Genetics

Amir Bahmani

Co-PI

School of Medicine

Genetics

Large language models (LLMs) like ChatGPT are tempting tools for students to use to complete various forms of assessments, from rhetorical writing to programming. Inspired by this problem, this scholarship team recently released DetectGPT, which uses an LLM to automatically detect its own outputs. While DetectGPT and related systems recently developed by OpenAI and Turnitin are promising steps towards automated detection of machine-generated text, standardized measurements of detector quality are missing, making comparison of detectors impossible and leaving educators in the dark about whether a detector is trustworthy. The research team proposes a new benchmark for machine-generated text detectors, addressing blind spots in existing evaluations. They will use this evaluation suite to develop the next generation of detection algorithms.

Name

Role

School

Department

Chelsea Finn

Main PI

School of Engineering

Computer Science

Christopher Manning

Co-PI

School of Engineering

Computer Science

This project aims to examine the potential of generative AI models in facilitating authentic problem-solving in science and engineering domains, and to determine the extent to which college students can learn to leverage AI to enhance their problem-solving practices and outcomes. The research team will also explore how science and engineering experts use ChatGPT to augment their problem-solving, which will lead to a framework of AI-human collaborative problem-solving practices. The research will have important implications for STEM education and how to prepare students for a future of human-AI collaboration.

Name

Role

School

Department

Carl Wieman

Main PI

School of Humanities and Sciences

Physics

Shima Salehi

Co-PI

Graduate School of Education

Graduate School of Education

Nick Haber

Co-PI

Graduate School of Education

Graduate School of Education

This project aims to use conversational AI and Virtual Reality (VR) to create interactive 3D avatars of medical virtual teaching assistants that can simulate real-world medical training in virtual environments. This team proposes MAI-TA, a medical conversational virtual agent that can supplement in-person teaching for personalized exploratory learning. Leveraging prior work on educational VR with anatomy photogrammetry scans, they will integrate OpenAI's GPT-3 to afford students a conversational way to explore digital anatomical specimens with customized guidance in a virtual lab setting. This project builds on previous research demonstrating that VR and digital anatomy labs can broaden access to medical training with under-represented and under-resourced learners.

Name

Role

School

Department

Sakti Srivastava

Main PI

School of Medicine

Surgery - Anatomy

Ken Salisbury

Co-PI

School of Engineering

Computer Science

Joel Sadler

Co-PI

School of Medicine

Surgery - Anatomy

This project uses generative models to engage students in the invaluable process of critical thinking and writing. The research team proposes deploying ChatGPT in their Stanford course ESF 17/17A "What Can You Do for Your Country", which asks students to read historically important texts about public service, from Kennedy's speech to Thucydides' "Pericles Funeral Oration," Lincoln's "Gettysburg Address," Frederick Douglass' "What to the Slave is the Fourth of July," and many others. Thus far, they have seen that generative models can help students better learn and articulate their ideas about public service. The team expects that building new pedagogical approaches and assessment including generative models will add pedagogical value.

Name

Role

School

Department

Russell Berman

Main PI

School of Humanities and Sciences

German

Ruth Starkman

Co-PI

Vice Provost for Undergraduate Education

Writing and Rhetoric