Recommendations as treatments

Joachims, Thorsten; London, Ben; Su, Yi; Swaminathan, Adith; Wang, Lequn

doi:10.1609/aimag.v42i3.18141

Cited by 8 publications

(9 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, note that a probe can be seen as an experimental trial as well, e.g. in the recommendations as treatments framework [5] products are seen as treatments that the recommender system chooses or not to administer to each of its users.…”

Section: (C) a Family Of Causal Link Prediction Tasksmentioning

confidence: 99%

“…Having this in mind, it is clear that independent contexts can be expected to hold by sampling training (interventional) data from distant parts of the graph. Lastly, having identical context (exogenous) distributions is a challenge often arising in causal inference [5]. A common way to accomplish it through experimental design is to stratify, i.e.…”

Section: (F) Proposition 411 In Practicementioning

confidence: 99%

See 1 more Smart Citation

Causal lifting and link prediction

Cotta,

Bevilacqua,

Ahmed

et al. 2023

Proc. R. Soc. A.

View full text Add to dashboard Cite

Existing causal models for link prediction assume an underlying set of inherent node factors—an innate characteristic defined at the node’s birth—that governs the causal evolution of links in the graph. In some causal tasks, however, link formation is path-dependent : the outcome of link interventions depends on existing links. Unfortunately, these existing causal methods are not designed for path-dependent link formation, as the cascading functional dependencies between links (arising from path dependence ) are either unidentifiable or require an impractical number of control variables. To overcome this, we develop the first causal model capable of dealing with path dependencies in link prediction. In this work, we introduce the concept of causal lifting, an invariance in causal models of independent interest that, on graphs, allows the identification of causal link prediction queries using limited interventional data. Further, we show how structural pairwise embeddings exhibit lower bias and correctly represent the task’s causal structure, as opposed to existing node embeddings, e.g. graph neural network node embeddings and matrix factorization. Finally, we validate our theoretical findings on three scenarios for causal link prediction tasks: knowledge base completion, covariance matrix estimation and consumer-product recommendations.

show abstract

Section: (C) a Family Of Causal Link Prediction Tasksmentioning

confidence: 99%

Section: (F) Proposition 411 In Practicementioning

confidence: 99%

Causal lifting and link prediction

Cotta,

Bevilacqua,

Ahmed

et al. 2023

Proc. R. Soc. A.

View full text Add to dashboard Cite

show abstract

“…To illustrate what is wrong with this seemingly reasonable idea, let us consider a simplified example of a news website that serves both left-and right-leaning users; this example is, in effect, borrowed from Singh and Joachims (2018) and Joachims et al (2021). Every day, there are some left-leaning new articles and some right-leaning new articles, and it is known that each user prefers the news articles that are closer to his/her own beliefs.…”

Section: Limitations Of the Current Approachmentioning

confidence: 99%

“…1.4 What is currently proposed to overcome this limitation A usual opinion in the recommender community is that the above limitation shows that we must go beyond utility and explicitly take fairness into account when selecting levels of exposure; see, e.g. Joachims et al (2021).…”

Section: User-oriented Websitesmentioning

confidence: 99%

See 1 more Smart Citation

How to make recommendation systems fair: an adequate utility-based approach

Tansuchat

Kosheleva

2022

AJEB

View full text Add to dashboard Cite

PurposeIn user-oriented websites, e.g. in news websites or in seller websites, it is important to take the user’s preferences into account when deciding which items to place in higher-exposure locations. The traditional approach to solving this problem, based on maximizing the average user utility, leads to unfair solutions, and this eventually hurts the company’s bottom line. Because of this, researchers have proposed complex schemes that explicitly add fairness to the formulation of this problem. But since utilities already describe human preferences, it is strange that it is necessary to add something beyond utilities.Design/methodology/approachIn this paper, the authors analyze the problem of selecting exposure level for different items from the viewpoint of decision theory, the basic theory underlying all our activities, including economic ones.FindingsThe authors show that a more adequate use of utilities, namely, taking into account that Nash’s bargaining solution is a proper way to make group decisions, not maximizing average utility, already leads to fair solutions.Originality/valueThe idea to apply Nash’s bargaining solution to the problem of assigning exposure level to different items is new, as well as the analysis that shows that this application restores the fairness, which is missing in the current solutions.

show abstract

Doubly Robust Estimator for Off-Policy Evaluation with Large Action Spaces

Shimizu,

Forastiere

2023

2023 IEEE Symposium Series on Computational Intelligence (SSCI)

View full text Add to dashboard Cite

Recommendations as treatments

Cited by 8 publications

References 66 publications

Causal lifting and link prediction

Causal lifting and link prediction

How to make recommendation systems fair: an adequate utility-based approach

Doubly Robust Estimator for Off-Policy Evaluation with Large Action Spaces

Contact Info

Product

Resources

About