CONSEQUENCES '24

Call for Contributions

Whilst early works framed the recommendation task as a prediction problem, there is a growing consensus that it is in fact a decision-making problem. Indeed: these systems take actions that have consequences. Nevertheless, often only short-term or myopic consequences are considered (e.g. "will this recommendation lead to a click?").

To reason about such consequences, possibly extending to longer-term consequences, we require a causal framework.

These consequences come in many forms: algorithms are often responsible for collecting training data for the next iteration, but algorithmic decisions also impact the utility that all stakeholders can get from participating on the platform. Indeed, the recommender systems research community is well aware that solely optimising for short-term engagement metrics is not always in the best interests of all stakeholders (such as item providers or the platform itself), but the research literature that jointly deals with these problems remains sparse.

The CONSEQUENCES ‘22 workshop on Causality, Counterfactuals, and Sequential Decision-Making focused on these topics from a general point-of-view, and was met with significant interest from both the academic and industry research communities. The CONSEQUENCES '23 workshop built upon this success whilst explicitly encouraging contributions that take into account multiple viewpoints and aim to understand, model and shape the consequences of recommendation decisions on all possible fronts. It is our belief that advances in the intersection of these fields foster progress in effective, efficient and fair use of logged data for both learning and evaluation, and can have strong impact on practical systems. As such, we believe the CONSEQUENCES '24 workshop will find an interested audience, and contributions from across industry and academia.

The CONSEQUENCES workshop series aims to bring a dedicated forum for learning and exchanging ideas. To this end, we welcome contributions from both academia and industry and bring together a growing community of researchers and practitioners interested in sequential decision making, reinforcement learning, offline and off-policy evaluation, batch policy learning, fairness in online platforms, as well as other related tasks, such as A/B testing.

The CONSEQUENCES workshop is an in-person event co-located with the Eighteenth ACM Conference on Recommender Systems in Bari, Italy. If necessary, arrangements will be made to allow speakers to present their accepted work remotely.

The workshop will accept contributions in the form of extended abstracts over three tracks:

1. Theoretical and Methodological Research

2. Benchmark, Datasets, and Software

3. Open Problems and Industry Experiences

We welcome both abstracts covering previously published work, as well as abstracts describing novel contributions or ongoing research. CONSEQUENCES '24 is a non-archival workshop to not obstruct future submissions to archival conferences and journals. 

List of Topics

Relevant topics to the workshop include, but are not limited to:

• Multi-armed and Contextual Bandits, and other policy-learning problems for recommender systems

• Learning, finding and designing metrics for long-term engagement and satisfaction

• Benchmarks, datasets, and software

• Causality in recommender systems

• Counterfactual inference for learning and evaluation of recommender systems

• Fairness and other long-term objectives in rankings and recommender systems

• Open problems and challenges in applications

Important dates

Submission deadline August 30th, 2024

Author notification September 13th, 2024

Camera-ready version deadline September 20th, 2024

CONSEQUENCES ‘24 October 14th, 2024

These dates consider in particular the synchronisation with the notification of acceptance in the main conference track, and the availability of early registration for the authors of accepted workshop papers.

Deadlines refer to 23:59 (11:59pm) in the AoE (Anywhere on Earth) time zone.

Submission instructions

Submissions occur via the central ACM RecSys EasyChair account: https://easychair.org/my/conference?conf=recsys2024workshops.

Formatting instructions

Submissions are limited to a single extended abstract with a length of at most two pages (excluding references) in the new double-column RecSys template. Authors can optionally include an appendix with further technical details to support the reviewing procedure.

Reviewing process

Papers will be reviewed following a single-blind process. All accepted papers will be presented as posters, selected papers will be presented as long talks.

Publication process

Accepted contributions will be made publicly available on the workshop website as non-archival reports, allowing future submissions to archival conferences and journals.
Contributions will be invited to present their work as an in-person or virtual talk or as a poster presentation, decisions about invitation will be based on content and expected availability of in-person presentations.

Organisers

• Olivier Jeunen, ShareChat

• Harrie Oosterhuis, Radboud University

• Yuta Saito, Department of Computer Science, Cornell University

• Flavian Vasile, Criteo

• Yixin Wang, University of Michigan