Online Talk | Manon Revel | 11 Mar 3pm
14 views
Skip to first unread message
Yara Kyrychenko
Mar 6, 2026, 10:08:23 AMMar 6
to Social Choice for AI Ethics and Safety
Dear all,
You are invited to join the next talk in the Cambridge SPSS talk series online on Wednesday 11 March 2026, 15:00-16:00 UK time.
Manon Revel (Google DeepMind) will speak about her research on representative ranking for online deliberation.
Title:
Representative Ranking for Deliberation in the Public Sphere
Abstract:
Online comment sections, such as those on news sites or social media, can foster informal public deliberation. However, this potential is often undermined by toxic or low-quality exchanges. To combat this, platforms increasingly leverage algorithms to facilitate higher-quality discussions, e.g., by using civility classifiers or forms of prosocial ranking. Yet, these interventions may also inadvertently reduce the visibility of legitimate viewpoints, undermining a key aspect of deliberation: representation of diverse views. This paper adopts the notion of justified representation (JR) from the social choice literature and incorporates a JR constraint into algorithms for ranking comments. We find that enforcing JR leads to greater inclusion of diverse viewpoints while still being compatible with optimizing for user engagement or other measures of conversational quality. This is joint work with Smitha Milli, Tyler Lu, Jamelle Watson-Daniels and Max Nickel.
Please contact me (yk...@cam.ac.uk) to request the videoconference login credentials.
Manon Revel (Google DeepMind) will speak about her research on representative ranking for online deliberation.
Title:
Representative Ranking for Deliberation in the Public Sphere
Abstract:
Online comment sections, such as those on news sites or social media, can foster informal public deliberation. However, this potential is often undermined by toxic or low-quality exchanges. To combat this, platforms increasingly leverage algorithms to facilitate higher-quality discussions, e.g., by using civility classifiers or forms of prosocial ranking. Yet, these interventions may also inadvertently reduce the visibility of legitimate viewpoints, undermining a key aspect of deliberation: representation of diverse views. This paper adopts the notion of justified representation (JR) from the social choice literature and incorporates a JR constraint into algorithms for ranking comments. We find that enforcing JR leads to greater inclusion of diverse viewpoints while still being compatible with optimizing for user engagement or other measures of conversational quality. This is joint work with Smitha Milli, Tyler Lu, Jamelle Watson-Daniels and Max Nickel.
Please contact me (yk...@cam.ac.uk) to request the videoconference login credentials.
Best,
Yara
_______________________________
Yara Kyrychenko
Social Decision-Making Lab, University of Cambridge
Reply all
Reply to author
Forward
0 new messages