Embedded Files
Algorithmic Recourse in ML
Online
Recording:
https://www.youtube.com/watch?v=TA64cfSiSEA&feature=youtu.be
A gathering of researchers and practitioners interested in the topic of algorithmic recourse and counterfactual/contrastive explanations in machine learning.
A gathering of researchers and practitioners interested in the topic of algorithmic recourse and counterfactual/contrastive explanations in machine learning.
The field of algorithmic recourse (incl'd counterfactual/contrastive explanations) has recently exploded with new papers covering technical, philosophical, psychological, and discussion-based perspectives.
To unify these recent developments, this short (1.5-2 hrs) get-together has been organized with the following agenda:
We received an overwhelming number of registrations, with 50+ attendees and 30+ groups interested in giving a short presentation. Given the limited time duration of this event (1.5-2 hrs), we are experimenting with a new format: following two invited talks by leaders in the field, we will have a longer discussion panel where all attendees are welcome to participate in both asking and answering questions (every attendee is a panelist!).
Given the expertise of attendees, and to sufficiently cover related topics, through both the invited talks and the panel, the gathering will focus on the following topics/questions:
To guide a fruitful discussion, and to get a map of everybody who is attending, please add your name, institution, research keywords, and 1 relevant paper to this list: https://bit.ly/3lGoJYd
Finally, this event does not require a NeurIPS registration, and the meeting will be recorded and streamed.
Best regards,Amir-Hossein Karimi
Amit Dhurandhar
Isabel ValeraKarthikeyan Shanmugam
To unify these recent developments, this short (1.5-2 hrs) get-together has been organized with the following agenda:
- Welcome remarks
- Invited talk 1 - Suresh Venkatasubramanian - A philosophical perspective on recourse
- Invited talk 2 - Ruth Byrne & Mark Keane - Counterfactuals in XAI: Human and Computational Studies
- Panel
- Closing remarks
We received an overwhelming number of registrations, with 50+ attendees and 30+ groups interested in giving a short presentation. Given the limited time duration of this event (1.5-2 hrs), we are experimenting with a new format: following two invited talks by leaders in the field, we will have a longer discussion panel where all attendees are welcome to participate in both asking and answering questions (every attendee is a panelist!).
Given the expertise of attendees, and to sufficiently cover related topics, through both the invited talks and the panel, the gathering will focus on the following topics/questions:
- What differences are there between “counterfactual” and “contrastive” explanations?
- How do counterfactual explanations relate to and complement other explainability methods?
- How does explainability interact with other ethical criteria such as fairness, privacy, strategic behavior, etc.?
- What are the biggest shortcomings and what do you see as the most important next steps?
- How do stakeholders best understand explanations in the context of AI/ML? What steps should we take as a community for the adoption of our methods in practice?
To guide a fruitful discussion, and to get a map of everybody who is attending, please add your name, institution, research keywords, and 1 relevant paper to this list: https://bit.ly/3lGoJYd
Finally, this event does not require a NeurIPS registration, and the meeting will be recorded and streamed.
Best regards,Amir-Hossein Karimi
Amit Dhurandhar
Isabel ValeraKarthikeyan Shanmugam
Page updated
Google Sites
Report abuse