Shaping the Future of Higher Education: R2D2 or HAL 9000

In academia, the increasing use of GenAI tools among administrators, students, and educators to ease administrative burdens, aid research efforts, and facilitate teaching and remote learning, among other things, challenges institutions of higher education to think of innovative ways to incorporate the benefits of GenAI tools into the learning environment while continuing to consider the potential negative consequences arising in this developing legal landscape.

Below, we outline key legal issues higher education institutions should keep in mind.

Intellectual Property Considerations

US copyright and patent laws currently do not afford protection to works or inventions created solely by GenAI. While some works, or elements thereof, may be eligible for intellectual property protection if they have a sufficient level of human authorship, what qualifies as “sufficient” is highly fact-specific and currently poorly defined. This ambiguity surrounding protection for GenAI-created or -influenced works could have significant implications for research institutions. Many universities obtain intellectual property rights to works of authorship and inventions produced or created using university grant funds or using university facilities, but these rights could be undermined if GenAI tools are used without a university’s knowledge or consent.

For example, a university would need to know about the use of GenAI before filing applications with the US Copyright Office or Patent Office so that its use could be properly disclosed or described. Among other things, universities could consider requiring faculty to disclose the use of GenAI in works or inventions in which the university will obtain rights or ownership.

Grant Writing and Academic Research

In an effort to regulate the widespread use of GenAI tools in academia, in particular in the fields of research and scholarship, some universities and research institutions have established guidelines for the ethical use of GenAI tools.

The Michigan Institute for Data Science at the University of Michigan, for example, permits the use of GenAI tools in drafting research papers and grant proposals, provided that the author include detailed citations regarding their use, and in some cases, include details about the specific prompts that the author used to generate the referenced information. Such guidelines can help discourage plagiarism and encourage transparency regarding the context and limitations of the ultimate work product.

Plagiarism and Discrimination

As the use of GenAI tools in academic settings becomes more prevalent, so too does the challenge of detecting and preventing plagiarism. Since the advent of GenAI, AI-based plagiarism has been on the rise and attempts to detect such infractions continue to face questions of reliability.

In some of the most alarming cases, the fallibility of GenAI detectors has raised concerns of discrimination. Detection technologies have been found to misclassify writing samples produced by non-native English speakers as GenAI generated works at higher rates, particularly in cases where the linguistic patterns and expressions of the author were limited, at least from an English language perspective. One study conducted by a professor at Stanford University found that as many as 60-98% of essays written by non-native English speakers for a Test of English as a Foreign Language (TOEFL) were erroneously flagged as GenAI generated essays.

Until detection technology becomes more reliable, it is important for institutions as both employers and educators to understand the limitations of such technologies and to consider the potential liabilities, including as related to claims of discrimination, associated with the use of these tools.

Admissions

Using GenAI tools may be beneficial in conducting a preliminary review of student applications in the admissions process — namely, ensuring that the more objective requirements of admission in a particular institution are met. While such tools may not be prudent to use to analyze the merit of essays since, as mentioned above, they have the potential to discriminate against well-qualified applicants, GenAI may still be extremely useful in determining whether applicants have plagiarized well-known works in their admission essays.

In order to work out any programmatic glitches, institutions may consider utilizing GenAI on a test basis for a year or two prior to implementation. This test period could allow higher education admissions officers the opportunity to tailor GenAI to the particular requirements of an institution, and, ultimately, institutions may be able to obtain higher yields and better retention.

Teaching

As GenAI tools have the potential to enable educators to conduct exhaustive research on a quick basis, they may be invaluable as teaching (and research) aids, enabling instructors to be more productive and providing enhanced educational opportunities for students.

However, one of the main problems identified with the use of GenAI continues to be the reliability of its data set — the internet. With the ongoing misuse of social media, widely reported inaccuracies, and questionable sources, the use of the internet has raised concerns with respect to the integrity of works produced by GenAI.

Institutions can ameliorate this concern by considering customizing their own data sets, such as tailoring GenAI tools to use the institution’s own data, other scholarly data, and reputable and independent sources, with proper attribution and payment, if applicable.

Remote Learning

One of the most popular uses for GenAI in the post-pandemic world has been to transform the learning environment. From dictation technologies to applications with the ability to generate lesson plans and provide supplemental tutoring, there is a strong case for using GenAI tools in educational settings to ease both administrative burdens on faculty and provide customizable and easily accessible learning environments for students.

However, the use of such tools also raises privacy concerns. Because adaptive tools, such as large language models (LLMs), are trained on data sets that often include personally identifiable information, users have limited, if any, control over their information’s use.

Before using GenAI tools, institutions should consider whether these tools comply with the privacy and data security standards required under state privacy laws to which the institutions are themselves subject, and whether their privacy protocols and protections are adequate to address these vulnerabilities and limit potential institution liability.

First Exam

As with any new product, GenAI should be tested and tailored given its current limitations. However, to not consider developing this promising tool now, recognizing its significant concerns, risks putting higher education institutions behind the learning curve.