The fear of AI
The focus of this post is the fears we have of AI technologies and what we might do about them. Having spoken to lots of librarians about AI, it is clear that many are concerned about the technologies but do not have time to address their fears in practical ways. Policy development obviously requires a lot of insight into the technologies in question and can be time consuming. There are hundreds of policy statements out there and we are in danger of not only being overwhelmed by the technologies themselves but by the policy statements as well.
An approach to developing policies
Professor Luciano Floridi, of the OII (soon to be at the Digital Ethics Centre at Yale) suggests in the third episode of the freely available University of Bologna online lecture series (time: 54:29) that many concerns are already being addressed by existing lists of principles for example the EU and, at a less legislative level, by the OECD. Commercial organisations could adopt these lists of principles but so could universities. In his work Floridi suggests that there are many overlaps and potential convergences which could make the drafting of guidance, regulations and policies easier. Floridi’s work on AI ethics is well worth considering as the basis for work on ethical policy in the higher education sector.
We can of course seek to develop our own policies, for example see this recent one from Bristol, but we could consider adopting existing highly informed policy statements or at least adapt them to our particular needs and circumstances. The OECD has created a comprehensive AI toolkit. It and related AI information can be found here. Could we adopt this approach and modify our policies to context specific form? The OECD approach is well informed by data and has been developed with input from a large panel of international experts. It is also worth taking a look at the OECD AI observatory dashboard which tracks AI policies from some 60 countries.
Policies for using AI in library work
Librarians are under considerable pressure and the workloads in libraries are not going down. Some have asked us about using AI to support the backlog in cataloguing. AI has the potential to help with some repetitive tasks (see Teaching an Algorithm How to Catalog a Book, Ernesto William De Luca et al*). Using these technologies for day to day library work will need to be underpinned with clear policies to regulate their use.
Influencing the wider environment
If we develop policies these will also have an impact on how AI develops in society because strong statements from organisations applies pressure to commercial providers to pay attention to the ethical development of their products.
* Computers 2021, 10(11), 155; https://doi.org/10.3390/computers10110155