Generative AI Hits Campus; is Campus Ready?
A snapshot of the university response to GPT-3 and ChatGPT, and a call to action
Educause is the largest association in the US focused on the use of information technology to advance higher education. At present, its membership includes more than 2,300 colleges, universities and 300 corporations. It has 16,500 active members and publishes on a wide range of topics of interest to the membership.
One of the association’s most recent publications is a detailed analysis of a poll of approximately 1,000 higher education stakeholders (C-level administrators, Director/manager, Staff, and Faculty) on attitudes toward the use of generative artificial intelligence tools specifically Generative Pre-trained Transformer-3 (GPT-3) based ChatGPT. The polling data points to a generally positive attitude or at least openness to the technology with 54% of respondents declaring themselves optimistic or very optimistic about its potential uses while only 12 % were classified as pessimistic or very pessimistic. But digging deeper into the poll results reveals much more institutional confusion and very serious faculty concerns about grading and student misuse of the technology than is readily apparent from the initial results.
While most respondents affirmed that they had heard of generative AI and 46% claimed to be very or extremely familiar with the technology, only 38% of the faculty, who will have the most responsibility for utilizing it in teaching, student grading and research, made the same claim. Faculty were also the most negative about its potential impact with 28% indicating that they felt pessimistic or very pessimistic about using the tool. However, the survey result that indicates broad underlying apprehension about the use of large language models (LLM) like GPT3/ChatGPT, is that the top concern of all respondents (75%) is cheating by students. An additional 60% had the related concern that content created by an LLM would be indistinguishable from human content; thereby thwarting efforts to discover and punish plagiarism.
At this stage, the university-level response can only be characterized as halting and uncertain. A majority (65%) of C-level leaders and managers indicated that they were discussing LLM’s with their colleagues not at all (10%), very little (15%) or somewhat (40%) with only 9% affirming they were deeply engaged in such discussions. Even more concerning is that when respondents were asked about the current and potential impact of LLM’s on thirteen key fields of institutional activity, majorities indicated that they saw no existing impact in any of those areas even including undergraduate teaching (45%) instructional technology (49%) and faculty research (51%). There is also a considerable amount of ignorance about how students were using LLMs with 55% indicating that they did not know. Finally, when respondents were asked about what policies related to LLMs were being developed at their institutions, the majority of respondents said that they didn’t know of any policy that was being developed, or that their institution is at a very early stage of discussion.
It seems obvious that generative AI LLMs will have a tremendous impact on all aspects of higher education, and equally obvious that the institutional response so far has been slow and tentative. The technology however is developing at a very rapid rate due to enormous investments by major tech companies and public interest in the technology is intense. By February 1st, 2023 ChatGPT had reached 100 million monthly active users making it the “fastest growing consumer application in history” according to UBS investment bank. Furthermore, GPT-4 will be available in a few months. GPT-4 will not encompass more than the 175 billion parameters contained in GPT-3 and should be better at discerning human intentions from prompts.
The potential impact of these new technologies on the university cannot be underestimated. It is imperative that institutions think through their integration and usage across disciplines as well as their administrative function now, or be overwhelmed. We will be doing subsequent posts exploring how universities can integrate these technologies in a measured, prepared manner.
To conclude, we can hardly do better than to quote from the Educause report itself:
As users quickly adopt generative AI for life and work, time is of the essence. The most productive immediate action stakeholders can take is to bridge institutional silos for focused discussion around the implications of generative AI for their communities.
Edwards, Benji, ChatGPT Sets Record For Fastest-growing User Base in History. Ars Technica February 1, 2023. Retrieved March 1, 2023.
Muscanell,Nicole and Robert, Jenay, Educause Quick Poll Results: Did ChatGPT Write This Report? Educause Research Notes, Emerging Technologies, and Trends February 14, 2023. Retrieved March 1, 2023.