I apologise - I'm not a museums person (I'm a former heritage specialist) but given the topic I thought I'd join in. I'm a final year Computing student with the OU and my final year project involves using machine learning to update historic environment records (HERs) direct from grey literature reports.

I suspect there are many reasons for not utilising AI - some of which have already been covered.

As already mentioned AI covers a very broad range from self driving cars down to machine learning classification. However even the machine learning classification end of the spectrum can be difficult to build & therefore somewhat resource hungry. I know there are projects that have been undertaken in archaeology with regards to text mining e.g. (Richards, Tudhope & Vlachidis 2015 Text Mining in Archaeology: Extracting Information from Archaeological Reports) but there were issues to do with retrieving the required data. For example Richards et al had difficulty with extracting the location because reports often make reference to other sites as comparators which can be confusing. My own experience is that is takes much more time and effort to build something usable than I had realised. I'm by no means an expert and more experienced developers probably could build something more quickly and efficiently but it is still an issue. I'd give more details of my experiences building an algorithm but I need to wait until after the project has been marked before I can share the results widely!

With resources being so tight I would assume that museums have to choose carefully what to invest in. A machine learning algorithm (or any other AI software) would need to do enough work to justify the expense spent on it. At the moment it may be cheaper to continue using humans, especially as there is the flexibility to use their labour for other tasks. For one museum to develop a classification tool that only they could use may not make economic sense - I suspect it would need to be an industry wide project to be more viable.

Whilst museums may not widely be using AI (in whatever form) now is the time to work out the ethics of using it in future. I have taken into account the ethics of developing my algorithm both for my client HER and the wider community. There is so much to consider - GDPR, job losses, copyright, system security to name but a few. As it is a university project it is perhaps not surprising that the university asked us to submit our ethical considerations early in the project. I don't know how ethics are handled in real world development situations in comparison - I will once I start working as a software engineer! However it would be no bad thing if museums were more aware of the ethical issues when commissioning work.

I've been fascinated by this topic so thanks for starting it! I may not work in heritage anymore but I love seeing how museums could make use of AI technologies.

all the best

Sarah Foxley

OU Computing and IT student

---

To unsubscribe from the MCG list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=MCG&A=1