What Happens When You Submit a Paper on AI Ethics to an AI Conference?

Well, we won’t name the conference (yet)! But it takes place this month, it’s a fairly standard ‘digital communications’ affair, with this year’s focus on AI (although not exclusively).

So we submit a paper on AI Ethics …

What happens then?

Well, the paper gets accepted.

Very quickly – which is a clue!

So, that’s nice.

Then we look at the review/feedback …

And we notice immediately that …

… the feedback has been written by AI!

(It’s very easy to spot. There’s an almost painful homogeneity to the language, structure and flow, there’s confusion/inconsistency as to whether an abstract or a paper is being reviewed, it suggests all the things that AI always suggests, and it matches similar feedback received from two other academics.)

Now, without immediately passing judgement on whether or not this is ethical, it’s probably worth noting two things:

1. This was a paper on AI ethics for heaven’s sake!
2. The conference hasn’t declared that it’s using AI for this purpose (either pre- or post- review)

This is the first draft slide of the presentation intended for the conference (unless we decide to withdraw it on ethical grounds) …

So IS this ethical?

Well, it’s perhaps up for debate (and we’ll probably continue the debate in the next post).

But maybe there are some obvious arguments for and against?

For:

• It’s efficient and it allowed for fast acceptance (although we don’t know yet what fraction of papers – if any – were rejected)
• It could be argued that this might avoid some forms of institutional bias (against women, minorities or young researchers, for example)

Against:

• The use of AI has not been declared
• The AI’s robustness has probably not been demonstrated for this purpose (How would it?)
• AI is being used to make decisions affecting humans without ethical discussion

Discuss!

(If we do decide to withdraw the paper, it might get published here)

To be continued …