Earlier in September Cabells published a blog about a controversy that had descended on the usually benign world of philosophy, where an article openly using AI had been provocatively published in a journal. The ensuing debate and decision by the journal to pretty much ban the use of AI from submissions has had wide repercussions and stimulated a more nuanced discussion about how and why AI can benefit research as a whole.
For example, the field of mathematics has been consumed by the idea of ‘dark knowledge’, triggered by a famous math problem that appears to have been solved by AI, but mathematicians aren’t quite sure how. This example of dark knowledge has led to a debate about research problems with the possibility of more societal impact than a mathematical proof, weighing the upsides of the benefits that could accrue from the use of AI in this way, to the downside that it is not human knowledge that is being extended.
Safety first
What these issues also highlight is the notion of safety and security with AI. In an article published recently in Science Business, the spotlight was placed on the disstrust some researchers feel towards AI regarding the risk that the work they share with large language models (LLMs) might be leaked or shared inappropriately somehow. For many authors, the risk of being scooped, or worse, sensitive data escaping from their labs is an existential threat to their careers and is something the guard fervently.
This is something Cabells understands intimately, and we ourselves have received numerous queries about our own AI tool, CompassAI. It is worthwhile stating at this juncture, therefore, that CompassAI – which can be used by authors to identify potential publications they might publish their research in using AI to match their work – does not ingest any data from users, nor does it track any information shared through the CompassAI portal. As such, users can have complete peace of mind that whatever they input into CompassAI will be done so securely and discreetly.
Global impact
While we can be confident that Cabells and other organizations will not leak user data, it is another thing to say that this will not happen for all academics. The problem here is three-fold: not only do we risk further polluting the scholarly record with academic research that is not ready or viable for sharing, we also infect LLMs with potential biases and erroneous information, and on top of that researchers will be less willing to make their research public to avoid some of these problems.
When it comes to questions of integrity and AI, it feels that researchers, publishers and information providers alike are feeling their way through a lot of new territory, and on the way, we find out a little bit more about who we are as individuals and organizations. It’s perhaps instructive for all parties to think about the integrity aspect first before worrying too much about what problems AI may represent.
