Andrew McStay

The Ethics of Emulated Empathy: Considering the Significance of Anthropic and Hume’s Collaboration! Dive into this interesting #blog by our AEGIS project! #AI. @digi_ad 📖rai.ac.uk/the-ethics-of-… 👉About the project- rai.ac.uk/research/impac…

A total pleasure to be a guest on the @BBCRadio4's 'Artificial Human', discussing whether AI can read emotions. The angle on the topic was excellent, as Chris, a listener, had called in asking whether AI can help him understand emotion expressions. bbc.co.uk/programmes/m00…

✨ #FENS – Chen Institute – NeuroLéman #SummerSchool: day 4! 🤝 The day kicked off with lectures by speakers @digi_ad and @GazzolaValeria, followed by a roundtable discussion, and wrapped up with a fun bowling dinner! 🎳 #FENSSchool! @_mameli_ @unil @NeuroLeman @ChenInstitute

There's right and easy criticism/concern here (privacy, mining, security, etc.) but, IMV, we shouldn't miss neurophenomenology (overlap of biology/experience). See here: academic.oup.com/book/55102/cha…

Ive been longing for someone to extend our work on ghostbots to synthetic avatars of the alive, like Abbatars, and here it is! Thanks ⁦@digi_ad⁩ ! link.springer.com/article/10.100…

New paper :) Here I argue that presence is a crucial factor in human-synthetic interaction, esp. in relation to ghostbots. It explains 'presence' and shows why it matters, esp. for those in governance. It gets weird, but there are v.simple recommendations. doi.org/10.1007/s10676…

@SimonadeHeer @mario_gug @APonceETUI Hi, perhaps, but I see wiggle room for a defence here. If one separates detection of states from expression (e.g. telesales), as per recital 18, I think there's a big problem in AIA. Problem is reliance on Ekman/physiognomy critiques. (Nb. I'd like to be wrong.)

@mario_gug @digi_ad @APonceETUI "Sounding grumpy" is an inference in itself. "Infer" was meant to be broader rather than restrictive. See recital 44: "AI systems intended to be used to *detect* the emotional state of individuals in situations related to the workplace and education should be prohibited."

#AI_Act to be noted: "emotional AI for monitoring could arguably discipline an employee if the AI says they sound grumpy on calls, just so long as there’s no inference that they are, in fact, grumpy." @digi_ad @APonceETUI * amp-theguardian-com.cdn.ampproject.org/c/s/amp.thegua…

@ChrisTMarsden Hi Chris, I don't think open access and my uni' doesn't subscribe. Got an open access link, or perhaps pop me an email with it?

(2) Citizens (n=2000+) had mixed views, seeing scope for safety, but also risks of bias (40% for, 27% don't know, and 33% uncomfortable). Qualitative work also mixed. Our view distills to "Don't": even for laudable goals (safety), risks are too high and the technology too shonky.

At the EAI Lab we find little evidence for effectiveness of emotional AI tech in a policing or security context, risk of bio-deterministic framings of criminality, deep skepticism from law enforcement (interviews), and high risk in trying to gauge emotion and/or human intent (1).

@ChrisTMarsden Thanks, missed this. I've seen a few trials by London Underground for anger, but not this one.

Most recent paper with @EmotionalAI colleagues 'On Manipulation by Emotional AI: UK Adults’ Views and Governance Implications' published here #biometrics #emotionprofiling #emotoys #manipulation frontiersin.org/articles/10.33…

This #ReviewoftheWeek centers on emotional #AI, evaluating the ethics of “contemporary technologies designed to emulate, interpret, and express empathy" in @digi_ad's "Automating Empathy" from @OUPAcademic: ow.ly/5EHG50RKogG #Empathy #AIEthics

Great to see @EmotionalAI and Project AEGIS represented by @exLaffer. I'm assuming Alex (in the middle) is just checking his talk notes ;)

UNESCO’s Ethics of AI team is recruiting! We're looking for 6 experts to contribute to cutting-edge projects on the ethical governance of AI across the following workstreams (please note the deadlines for applications are 16-17 April):

📢Exciting news! A new white paper titled “AI FRINGE PERSPECTIVES’ is now published, written by a consortium drawn from the attendees of @AISummitFringe with @responsibleaiuk ensuring a diversity of views garnered from the academic community. Download now rai.ac.uk/publications

And given the London Underground tests with emotion/intention recognition, my 2023 views for free here in book form! academic.oup.com/book/55102

As it stands, we are yet to see any emotion AI technology develop in a way that satisfies data protection requirements, and have more general questions about proportionality, fairness and transparency in this area. (October 2022) ico.org.uk/about-the-ico/…

Many problems, but detecting "aggressive" behaviour is v.notable. Esp' a problem for folk with dark skin (black people with neutral expressions more likely to be labeled angry). ICO got ahead of this issue (we inputted) so time to put that policy to work. wired.com/story/london-u…