Online Certification of Preference-Based Fairness for Personalized Recommender Systems

Authors: Virginie Do, Sam Corbett-Davies, Jamal Atif, Nicolas Usunier6532-6540

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also study the tradeoffs achieved on real-world recommendation datasets. ... In Sec. 5, we investigate the trade-offs achieved on real-world datasets. ... We present experiments describing sources of envy (Sec. 5.1) and evaluating the auditing algorithm OCEF on two recommendation tasks (Sec. 5.2).
Researcher Affiliation Collaboration Virginie Do1,2, Sam Corbett-Davies2, Jamal Atif1, Nicolas Usunier2 1LAMSADE, Universit e PSL, Universit e Paris Dauphine, CNRS, France 2Meta AI
Pseudocode Yes Algorithm 1: OCEF algorithm. ... Algorithm 2: AUDIT algorithm.
Open Source Code No The paper does not provide a direct link to open-source code for the methodology described, nor does it state that the code is being released.
Open Datasets Yes We create a music recommendation task based on the Last.fm dataset from Cantador et al. (2011)... We also address movie recommendation with the Movie Lens-1M dataset (Harper and Konstan 2015)
Dataset Splits Yes the simulated recommender system estimates relevance scores using low-rank matrix completion (Bell and Sejnowski 1995) on a training sample of 20% of the ground truth preferences
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU models, CPU types) used for running its experiments.
Software Dependencies No Using the Python library Implicit: https://github.com/benfred/ implicit (MIT License). This mentions a library but does not specify its version number, nor does it list other software dependencies with versions.
Experiment Setup Yes Recommendations are given by a fixed-temperature softmax policy over the predicted scores. ... We vary the number of latent factors of the matrix completion model and evaluate a softmax policy with inverse temperature set to 5. ... matrix completion with 48 latent factors. ... inverse temperature equal to 5 or 10.