Don’t Ban Artificial Intelligence From Academia. Reference It!
Advances in AI text generation pose a significant threat to universities. It is hard to detect plagiarism when the submitted work is original. A similar issue already exists with contract cheating. Students can go to dodgey, scam ridden sites and pay someone to write their assignments for them. But while these “essay mills” have been made illegal in several countries, the same response would be foolish for AI text generation. This is because AI offers clear efficiency gains for many professional writers.
Indeed AI companies market their services to the corporate world precisely for this reason. Even for serious academic research, AI will continue to offer key productivity gains. Imagine preliminary scans of archives that generate summaries for historians. This would save hours of time. Or, if all the data has been compiled, rough drafts of scientific papers could be generated. Whenever time is saved, already expensive research will reap greater rewards.
The only issue concerns transparancy, but this can be fixed with new referencing standards for AI. Of course, referencing AI is not as intuitive as might be assumed. It requires a particular understanding of the purpose of referencing. To some, referencing a source means flagging that a particular idea was the creation of some other person or group. If this is the purpose of referencing, then it doesn’t necessarily make sense to reference AI. Why? Because AI is based on a reading of millions of texts. In a sense it is the creation, in part, of all of us.
Fortunately, this is a silly way of thinking about referencing and authorship. For we are all like AI. We have all had our thinking shaped through “deep learning” with millions of “texts”, and countless other social experiences. The idea that the author of a particular text was the sole creator of its ideas has never made sense. So what, then, is the point of referencing?
- It shows who did the most immediate work (research and writing) of creating the particular text.
- It helps readers to track down the original sources.
Neither of these two purposes makes any claim about the extent to which an individual or a group is the sole creator or “owners” of a given idea. These points serve only to show who did the immediate work of research and writing to generate a given text, and where that text can be found.
According to this conception of the purpose of referencing, there is no reason why AI cannot be referenced. For example, referencing would help readers understand that a given percentage of the text was generated by AI. They would then understand who was responsible for the immediate task of producing a given text, including both research and writing, and they would, potentially, be able to reproduce similar results.
Accomodating AI in this way would allow any benefits of the technology to be used transparently and fairly. It would also force people to think more deeply about the authorship of research and writing, of the authorship also of our very lives.
