Dagstuhl: Engineering Moral Agents

Matryoshka, the Boss and a couple of other academics organised a Dagstuhl Workshop on "Engineering Moral Agents" in early June. I've written about Dagstuhl several times before so I won't re-iterate my descriptions of the place.

This workshop had quite a distinct feel to it, I suspect because they had managed to invite a fairly hefty number of philosophers to counter-balance the computer scientists. This had a number of interesting side effects. Philosophers love discussing almost anything in depth and at length. By the end of the second day we had had (unsurprisingly) a long discussion about the EU Referendum with a particular emphasis on a game theoretic account of how the EU should behave as a "rational actor" in the event of Brexit; what did we mean by a robot; was my dishwasher autonomous (given it can "choose" to refuse to wash the dishes if it thinks there is a danger of a flood); and whether Religion and Philosophy should be taught as part of the same subject in schools - to be honest it felt a little like a certain portrayal of undergraduate life which I felt I rarely experienced (most of the in depth conversations I recall from my undergraduate days were about things like whether sex had any place being mentioned in Doctor Who novels).

The Wednesday afternoon excursion was put to a popular vote revealing a distinct lack of enthusiasm for Trier in the rain, and a rather more surprising lack of enthusiasm for the Town that Exists to Sell Wine to Tourists. On the Wednesday morning we split into two working groups one, dominated by philosophers, ambitiously decided they were going to formalise ethical reasoning the other (containing, interestingly, all the women at the workshop plus a few others, including our token doom-sayer*) wanted to discuss the implementation of ethics. A side effect of this was that each group decided to organise its own walking tour into the German forest. I somehow ended up in charge of ours, I think because I had been to the local lake twice before and so, at least in theory, had some idea of the way there. I also went to reception and got them to give me a map and directions to supplement our numerous mobile phones. I'm sure you've be pleased to know that not only did I not loose any computer scientists or philosophers in the German forest, but even managed to find an extra computer scientist (from the planar graph workshop that was also taking place), her husband and baby and conduct them safely back to the schloss.

Matryoshka, the Boss and I have already submitted one paper as a result of the Dagstuhl (trying to model reposting behaviour across different social media services, as part of Matryoshka's agenda to fix the internet). Admittedly we decided to do this because I had already decided to attend the relevant workshop so it seemed like a good idea to at least submit a paper to it. Our working group is also putting together a position paper covering our discussions. Matryoshka and I were a little sceptical about this having experienced "we must publish our working group discussions" before, however on this occasion people are actually writing stuff, so it looks more hopeful. I agreed to start curating a set of examples of interesting ethical dilemmas (or at least cases where the ethical reasoning required wasn't entirely trivial) which would hopefully provide a useful benchmark resource for people interested in implementing ethical reasoning. I also had some useful discussions with one of the implementors of the major (if not the only) computational ethical reasoning system currently in serious existence about approaches to verifying aspects of their system which may also lead to some kind of publication. Taken together, that makes this a pretty productive Dagstuhl.

*we instantly implemented a rule which involved him paying a forfeit every time he said "super-intelligence".