from the the-speed-of-change dept

A key role universities play in society is providing knowledge that can guide public action. Moreover, public universities have a legal duty to serve the public. We expect researchers in universities to  produce and disseminate knowledge that can ultimately advance human flourishing and meet the informational needs of lawmakers, civil society, and the public generally. Yet too many schools are currently failing to meet this expectation in key domains. Nearly four years after the launch of ChatGPT sparked humanity’s most pressing technological moment, many public stakeholders are left with more questions than answers about how to navigate this AI moment. 

Universities have only just begun to enable scholars and students to contribute to the massive task of providing adequate knowledge to guide societal action with respect to AI. This delay–perhaps forgivable for a brief period of time given that universities will never move at the speed of startups–is eroding the capacity and, as a result, legitimacy of these critical institutions. The longer that universities operate as if society were not undergoing fundamental change, the faster that faculty members will depart for other opportunities (e.g., in industry), potential and current students will opt for alternative paths to education, and stakeholders will decrease or even withhold ongoing financial support. 

Other individuals and more nimble organizations have rushed to fill the informational vacuum. AI labs, research institutes, and partisan entities now play a heavy role in shaping how the public and elected officials think about this moment. However, these parties–no matter how well-intentioned–have biases that may skew public discourse and lack the legitimacy of institutions like universities, which are built to shield researchers from major conflicts of interest. 

The exodus of talent from many universities is partially outside of their hands. Federal and state lawmakers often control the resources available to public institutions. They have yet to equip universities with the resources and tools required to assist with navigating this AI transition. In some cases, lawmakers have actively diminished support for higher education. The foreseeable result is that many professors cannot access leading AI models and many students lack AI training and literacy. These decisions have produced an academic community that is technically challenged in its ability to guide society through the myriad questions posed by AI. 

Public officials–to varying degrees and levels of specificity–flag for universities what studies, projects, and events would help meet public informational needs, but universities may not prioritize that work. For example, agencies sometimes issue requests for information–invitations for the public, especially individuals with relevant expertise–that do not result in meaningful engagement. At the federal level, agencies often receive fewer than ten comments on critical regulatory questions. More targeted outreach programs to scholars could provide agencies with more information as well as give academics a chance to share their perspectives. Relatedly, when legislative committees at the state level hold hearings on topics closely studied by a nearby university, they often opt to invite politically connected witnesses rather than professors. This preference may make for good politics but it leaves academics on the sidelines. A small amount of effort by these state actors could go a long way toward helping academics shift their research priorities. 

In this way, external actors may be partly to blame for the increasing irrelevance of higher education in AI governance conversations. But universities should also take some responsibility. Even if professors were equipped with the latest AI, some would likely fail to use it–preferring to stick to the syllabus they’ve taught from for decades and investigate the topics that they find most interesting, regardless of societal relevance. If professors persist in the belief that AI is something they can ignore and act as though they have no responsibility to monitor and contribute to the public knowledge, there’s little reason for such public funds to go toward AI tools at universities nor to improve outreach to academics to assist with AI-related questions.

Academics are not passive players in decisions to cut funds and reduce support. Across disciplines, academics have maintained incentive structures that permit many scholars to act as though it’s still 2022. In response, some academics who  appreciate the scale, speed, and societal significance of AI have left the Ivory Tower for the Silicon Tower, where the pay is higher, the tools are more sophisticated, and the culture rewards those who dive into the most important inquiries. 

Given these challenges, what can be done? This essay provides an initial call for reform, to which we hope others will also contribute. Questions of how to resolve the compensation gulf between industry and academia are left for another time. Universities–including faculty, staff, administrators, and related organizations and networks–can still make progress on establishing the conditions for increased and expedited AI research by those who opt to stay in academia. These reforms should satisfy several foundational principles, such as protecting academic freedom, creating meaningful opportunities for student participation, and addressing issues with particular importance to the university’s surrounding community. 

Again, this is a starting point. We invite academics and others to share this call for reform, add to it, and amend it; what we should not do is carry on as normal. Instead, we call for the vast majority of scholars to substantially re-orient their research, writing, teaching and service around AI. 

We think that our legislators, civil society organizations, and neighbors would be better off with  too much analysis of AI, rather than too little. We’re currently living in a world with too little analysis and it’s far from satisfactory. Not every scholar should pivot to addressing AI directly, but the majority of departments should be able to point to how their collective expertise contributes to a better societal response to this moment. A surge in focus and expenditure on AI need not last forever but is highly needed at the moment.

Recommendation #1: Identify and Prioritize Societally Important Research Questions

Professional organizations like the Association of American Law Schools (“AALS”), which represents law schools and law professors, should establish and support shared services that collect consequential questions about AI from lawmakers and AI stakeholders, such as frontier AI labs and research entities, and connect them with relevant academic communities. Staff at these organizations have extensive knowledge of the communities that exist within the discipline as well as how best to share research, advice, and engagement opportunities (as explored below). 

Additionally, participating universities can and should set up means to receive AI-related requests to provide yet another avenue for scholars to learn about how they can contribute. Imagine, for example, if universities designated a “legislative research liaison” who had the sole purpose of working with the legislature to specify their knowledge gaps. (It is worth noting that some universities already have government affairs teams that commonly interact with legislators, albeit not necessarily for this purpose.)

Of course, individuals, organizations, and governing bodies can also share their informational needs with the Ivory Tower. This need not be a costly endeavor. For instance, state legislative committees could start their next legislative session with a hearing that culminates in a research agenda that is published and sent directly to all universities. 

Recommendation #2: Enable The Necessary Resources for Timely and Thorough Research on AI

State and federal officials, as well as AI stakeholders seeking expedited expert analysis of complex questions, should ensure that participating scholars have the resources to offer valuable insights. As it stands, universities have far too few tokens, limited access to frontier models (sometimes because of their own outdated procurement standards and IT policies), and, in some cases, minimal understanding of the latest developments in AI (such as the existing and near-future capabilities of AI). 

Government officials counting on scholars to help them in their work ought to invest in their capacity to do so. AI research is not cheap. Universities are struggling to provide meaningful access to faculty, staff, and students given the high costs of many AI systems and the variability of usage-based expenses. Rather than wait for universities to procure such access, some professors are trading university roles for the lab gigs–joining labs where they know that they will be able to use the latest and greatest AI. Legislatures have yet to make the sort of sizable investments in universities required to meet the demand for AI access among experts nor to compensate experts enough to keep them on campus. In fiscal year 2025, the average state appropriation for higher education amounted to just a 2 percent increase when adjusted for inflation. 

AI labs likewise counting on scholars to assist with readying society for their transformative technology ought to lower barriers to academic investigation. While labs have offered some universities financial breaks on their tools and permitted some degree of support to researchers, there’s substantial room for improvement. Labs, including Google and Anthropic, are sitting on vast troves of user data that can inform everything from how AI may impact the labor market to the extent to which individuals are becoming reliant on AI tools for sensitive tasks. Universities could unleash their extensive interdisciplinary experts on such data if labs facilitated access. Labs may have justifiable reasons for holding that information close to the chest. In the wrong hands, data related to user habits, interests, and goals could lead to problematic outcomes. That said, Anthropic, OpenAI, and others cannot at once stress their concern about society’s readiness for transformative AI while keeping the information necessary to assist with that effort locked away. 

Recommendation #3: Reward Contributions to Societally Relevant Work on AI

Academics may be hesitant to alter their research agenda and planned publication goals if doing so is likely to block professional progress. Tenure requirements are often tied to adherence to the existing, protracted processes. What’s more, even those requirements are completed on paper–i.e. publication of a certain number of articles, tenure decisions are commonly shaped by norms and biases that further obligate applicants to invest in the status quo. For example, in the discipline of economics, tenure may be especially likely to be awarded if the applicant has published in one of five journals. 

Alternative contributions to the discipline, especially those done in response to demand for input on AI questions, should clearly and meaningfully inform tenure, compensation, professional recognition, and perhaps university obligations, such as committee assignments. In addition to rewarding traditional research, universities eager to meet this AI moment should formalize recognition of the following: 

• public engagement research (albeit likely in a truncated form and via more accessible outlets),
• advice (testifying before legislative bodies, responding directly to government requests for information and the equivalent from labs), and 
• engagement (meeting with legislators, regulators, and AI company policy staff, teaching community courses, serving as an expert at community forums).

Accounting for these contributions to a moment of informational demand would return universities to their historical focus on knowledge creation and dissemination. Rep. Jordan Morrill, author of the eponymous act that created the nation’s land grant universities, introduced the legislation with the following vision: “Let us have such colleges as may rightfully claim the authority of teachers to announce facts, fix laws, and scatter broadcast that knowledge which will provide useful in building up a great nation.” That guiding principle informed subsequent legislation. The Hatch Act, enacted in 1887, set up “agricultural experimentation systems” at universities in each state for the purpose of:

acquiring and diffusing among the people of the United States useful and practical information on subjects connected with agriculture, and to promote scientific investigation and experiment respecting the principles and application of agricultural science.

These academic entities were expected to “conduct original research or verify experiments” related to timely agricultural questions and to produce “reports of progress . . . at least once in three months, one copy of which shall be sent to each newspaper in the states in which they are respectively located, and to such individuals actually engaged in farming as may request the same.” Notably, these reports were to be sent via the “mails of the United States” at no charge to the recipient.

Realization of Morrill’s vision and replication (with some adaptation) of the Hatch Act’s provisions in modern times would transform modern universities, many of which have serially drifted from their respective missions. 

Recommendation #4: Accelerate Publication and Delivery to Stakeholders

Academic research tends to move at the speed of their respective research journals, which are rarely known for their haste. As of 2013, the publication cycle–from submission to public distribution–in business and economic journals often took 18 months; physics, comparatively moving at hyper-speed, averaged 9 months for that cycle. While some degree of delay is unquestionably necessary to probe and refine research, many scholars would conclude that the time between paper submission and publication could be shortened with minimal or, given the speed of AI, perhaps even justifiable losses to quality. 

Rather than tolerate continued reliance on those processes, domains should identify suitable alternatives that may allow for sufficient review while not unnecessarily delaying release. While many pre-print avenues exist, such as ArXiv and SSRN, it may nevertheless still be the case that the existing norms around length and formatting standards still slow a scholar’s dissemination of information that may be especially valuable to AI stakeholders. For example, it may be the case that in certain scenarios the most valuable thing to state or federal officials is a wide-range of white papers or even two-page memos from a range of practitioners. Such memos would substantially depart from the traditional research paper in just about every domain yet have an incredible impact on a pressing policy matter. 

Minimally, disciplines should commit to making their research outputs open source when possible. Academic research that lives behind a paywall will inevitably fall short of its potential to contribute to public discourse. Yet, in many disciplines only a fraction of articles are open source. Theoretically, authors of paywalled pieces may share shorter, more accessible versions through alternative means. Still, it’s antithetical to the mission of universities–especially public universities–to impose gates on knowledge. The need for open knowledge is all the greater with respect to research that addresses timely matters, such as AI. Furthermore, the materials, data, and analysis code underlying empirical research (when relevant) should be made publicly available and be as transparent as possible. By adhering to this principle of open science, empirical fields can move faster, providing opportunities to ascertain which findings are reproducible and replicable more quickly.

Conclusion 

Universities have obligations to help support the societies in which they are embedded. Novel problems and possibilities presented by this AI moment undoubtedly call for substantial assistance from our educational institutions. The call has not been answered. In fact, many of the individuals best suited to assist universities at this time of public need are leaving. The public ought to demand more–from universities, from state and federal government officials, and from AI labs. 

Universities can and should change their priorities and practices to better serve their external stakeholders. Different universities will adopt different approaches. Some will be more successful than others. This letter is meant to spark conversations that lead to tangible reforms,, not to advance a single best approach. Each discipline and institution may have varied answers but we think they all ought to center on a contemporary version of what Walter Murray, first president of the University of Saskatchewan, wrote in 1908: 

What is the sphere of the university? Its watchword is service—service of the state in the things that make for happiness and virtue as well as the things that make for wealth. No form of that service is too mean or too exalted for the university. It is as fitting for the university, through correspondence classes, extension courses, supervision of farmers’ clubs, traveling libraries, women’s institutes or musical tests to place within the reach of the solitary student, the distant townsman, the farmer in his hours of leisure or the mothers and daughters in the home the opportunities for adding to their stores of knowledge and enjoyment, as it is that the university should foster researches into the properties of radium or the causes and cure of swamp fever.

It is an obligation for universities to serve the public. The need is especially high in this time of drastic technological change, with AI adoption happening faster than the adoption of any other technology in human history. We hope universities will embrace the recommendations in this essay, responding more effectively to the AI moment. 

Filed Under: academia, ai, research, universities