25 Research Papers That Study GenAI Business, Social Implications
A new compilation of work offers analysis and insights from top MIT Sloan scholars about AI’s progress and perils.
By Sara Brown, courtesy of MIT Sloan
In response to the rapid rise of generative artificial intelligence, in the fall of 2023 MIT president Sally Kornbluth and provost Cynthia Barnhart issued a call for research proposals related to how the technology will transform people’s lives and work.
The result is a new open-access collection of 25 research papers that provide road maps, policy recommendations, and calls for action about generative AI from MIT experts. Topics range from using generative AI in education, manufacturing, and drug development to how generative AI will affect inequality and music discovery.
The research papers are being widely shared as quickly as possible because generative AI is evolving at a rapid pace
and they could “serve as a springboard for further research, study and conversation about how we as a society can build a successful AI future,” Kornbluth said in an introduction to the collection. The papers are all works in progress and may be further developed. They have not been formally peer-reviewed.
MIT Sloan faculty members and researchers contributed to the following research projects.
Engineering and manufacturing
This paper examines the barriers, risks, and potential rewards of using generative AI for design and manufacturing. MIT Institute professor Daron Acemoglu, MIT Sloan economist Simon Johnson, and co-authors interviewed manufacturing experts and industry leaders to find weaknesses that need to be addressed by the next generation of generative AI tools.
Implementing Generative AI in U.S. Hospital Systems
AI can transform health care, but there are often challenges when new technologies are introduced into clinical settings.
In this paper, the researchers, including MIT Sloan professors Kate Kellogg and Retsef Levi, look at the difference between deploying narrow, or traditional, AI and generative AI in health care systems and how challenges associated with both technologies inform where AI might be most effective.
Work and productivity
The Impact of Generative AI on Labor Market Matching
MIT Sloan PhD student Justin Kaashoek and MIT Sloan professors Manish Raghavan and John J. Horton discuss areas where generative AI might appear in the labor market, including AI-generated cover letters, resumes, and job postings. The researchers examine the risks and benefits of these uses and identify ways to mitigate risks while promoting the benefits of AI.
The Productivity Effects of Generative AI: Evidence From a Field Experiment With GitHub Copilot
Two field experiments found evidence that software developers at Microsoft and Accenture who were given access to an AI-based coding assistant became more productive. The researchers, including MIT Sloan professor Mert Demirer, indicated that their findings are preliminary and that the team is still in the process of collecting additional data.
Bringing Worker Voice Into Generative AI
Input from workers can increase the likelihood that organizations use generative AI tools effectively and that workers’ job quality improves.
The researchers, including MIT Sloan professors Thomas A. Kochan and Emilio J. Castilla and MIT Sloan MBA candidate Ben Likis, identified ways to bring workers’ voices into the development and use of generative AI.
Practical applications
Generative AI From Theory to Practice: A Case Study of Financial Advice
Researchers, including MIT Sloan professor Andrew W. Lo, looked at the most pressing issues facing the adoption of large language models. Using financial advice as a test for determining the shortcomings of LLMs, the researchers suggest ways to improve generative AI in general.
Labeling AI-Generated Content: Promises, Perils, and Future Directions
How should policymakers, platforms, and practitioners decide how to label AI-generated content? MIT Sloan professor David G. Rand and other researchers outlined two goals for labels: communicating whether a piece of content was created or edited using AI, and diminishing the likelihood that content misleads or deceives its viewers. They found that under certain conditions, labeling can decrease individuals’ likelihood of believing or engaging with misleading AI-generated images.
Social implications
Data Authenticity, Consent, and Provenance for AI Are All Broken: What Will It Take to Fix Them?
New AI capabilities rely on massive, widely sourced, and under-documented training data. In their paper, the researchers, including MIT professor Sandy Pentland, identify the shortcomings of existing tools for data authenticity and consent and argue for new universal data provenance standards.
Generative AI for Pro-Democracy Platforms
Generative AI has the potential to promote civil discourse, gather citizen input on policy questions, and strengthen democratic practices. A research team that includes Pentland presents
a framework to help policymakers, technologists, and the public assess potential opportunities and risks when incorporating generative AI into online platforms for discussion and deliberation.
Generative AI and the Future of Inequality
Many fear that generative AI could lead to more inequality and even mass unemployment. MIT Sloan professor Nathan Wilmers summarizes existing research about how generative AI will likely affect the labor market and inequality. In particular, Wilmers looks at how the effects of generative AI will likely differ from previous technologies.
Originally published at https://mitsloan.mit.edu on April 16, 2024.
