#AAAI2023 workshops round-up 1: AI for credible elections, and accountable human-centric AI

As a part of the thirty seventh AAAI Convention on Synthetic Intelligence (AAAI2023), 32 completely different workshops had been held, masking a variety of various AI subjects. We hear from the organisers of two of the workshops, who inform us their key takeaways from their respective occasions.

AI for Credible Elections: A Name To Motion with Trusted AI

Organisers: Biplav Srivastava, Anita Nikolich, Andrea Hickerson, Tarmo Koppel, Sachindra Joshi, Chris Dawes.

This workshop examined the challenges of credible elections globally in an educational setting with apolitical dialogue of serious points. The important thing takeways from the workshop had been as follows:

How a expertise is designed and the way points round their utilization are dealt with can have an effect on voters’ belief in them. Digital voting has the issue of transparency and a paper path helps mitigate it considerably.
The completely different jurisdictions within the US have plenty of freedom in organizing elections, however in consequence, additionally chaos from that. It’s an open query whether or not that is fascinating for a reputable election.
AI can particularly assist elections by disseminating official data (e.g., about candidates, electoral course of and candidates) personalised to a voter’s cognitive wants at scale, of their language and format. When AI helps forestall mis- and dis-information, the influence might be far reaching.

You possibly can learn an extended abstract of the workshop right here.

R2HCAI: Illustration Studying for Accountable Human-Centric AI

Organisers: Ahmad Beirami, Ali Etemad, Asma Ghandeharioun, Luyang Liu, Ninareh Mehrabi, Pritam Sarkar.

The AAAI 2023 Workshop on Illustration Studying for Accountable Human-Centric AI (R2HCAI) introduced collectively researchers who’re broadly keen on illustration studying for accountable human-centric AI. The aim of the workshop was to facilitate the event and adoption of AI techniques that may improve, increase, and enhance the standard of human life.
We had six inspiring invited talks from famend researchers that lined a variety of analysis within the discipline of accountable human-centric AI. Marzyeh Ghassemi gave a chat on designing machine studying processes for equitable well being techniques, whereas Daniel Ruckert shared their latest work on human-centered AI for medical imaging. Kathy Meier-Hellstern shared a framework for accountable AI for big fashions, and Jacob Andreas introduced their analysis in the direction of pure language supervision. Hima Lakkaraju gave a chat on “Bringing Order to Chaos: Probing the Disagreement Drawback in XAI”. Lastly, Deepak Pathak shared their analysis on how robots can study from human movies by watching, working towards, and enhancing.
Beneath the moderation of Ali Etemad and Ninareh Mehrabi, Jacob Andreas, Kathy Meier-Hellstern, Deepak Pathak, and Daniel Rückert there was an insightful panel dialogue on illustration studying for accountable human-centric AI, together with a dialogue across the accountable AI points of highly effective generative fashions.
We congratulate Marwa Abdulhai, Clément Crepy, Dasha Valter, John Canny, and Natasha Jaques on successful one of the best paper award for his or her paper “Ethical Foundations of Massive Language Fashions”. This paper makes use of Ethical Basis Idea to investigate whether or not standard LLMs have acquired a bias in the direction of a specific set of ethical values.
We acquired 46 submissions from 219 authors, of which 30 papers had been accepted. 69 program committee members had been nominated by workshop chairs, authors and thru self nominations. There have been additionally 32 space chairs nominated by workshop chairs. To maximise alignment between paper subjects and reviewer experience, we gathered paper biddings from this system committee and space chairs and assigned papers to reviewers accordingly.

You possibly can see the complete record of workshops that occurred at AAAI 2023 right here.

tags: AAAI2023