INSPIRE

Not Unbiased: The Implications of Human-Algorithm Interaction on Training Data and Algorithm Performance

This project is supported by the National Science Foundation grant NSF IIS-1549981.

Introduction

This INSPIRE award is partially funded by the Information Integration and Informatics program in the Division of Information and Intelligent Systems in the Directorate for Computer & Information Science & Engineering, the Perception, Action & Cognition program in the Division of Behavioral and Cognitive Sciences in the Directorate for Social, Behavioral & Economic Sciences, and the Office of Integrative Activities in the Office of the Director.

One of the most common uses of machine learning is to learn to replicate human decisions, a common example is recommender systems. In these systems, computers are trained to replicate the recommendation a collaboration of hundreds or thousands of humans would give, if that were possible. Most of the data used to train these systems are not from a controlled random sample, but are obtained from users based on outputs of algorithms (e.g., which search engine results do users click on?), which introduces bias into the process and ultimately impacts the quality of the results. This project addresses this problem by examining how the human decision process that creates these data in the first place is affected by the data coming from machine algorithms, how this in turn impacts the algorithms themselves, and how to ultimately adjust for human bias in the machine learning process. Specific areas tackled are filtering (e.g., web search) and recommender systems. The deep research into how the human decision process affects machine learning, and how machine learning impacts the human decision process, can provide significant advances in the accuracy and utility of systems using machine learning.

The project builds on analysis of machine learning algorithms based on Hidden Markov Models (HMMs). The formal analysis initially looks at “blind spots” – the impact of bias from users not getting complete (or a random sample) of data. Further analysis will be based on the outcome of two human experiments: Two category recommendation (labeling items, with items to be labelled chosen by random, active learning, and filter-based algorithms), and movie recommendation. The results will be used to develop improved machine learning approaches based on antidotes (altering learned models to reduce bias) and reactive learning (active learning that takes into account the human and machine biases). The PIs also have plans to capitalize on the lessons learned by providing examples of the use of cognitive science in a Web Mining course, and of the impact of machine learning in Data Science for Psychologists courses

People

Faculty Investigators

Olfa Nasraoui – Principal Investigator

Patrick Shafto – Co-Principal Investigator

Graduate Students

Mahsa Badami

Wenlong Sun

Sami Khenissi

Mariem Boujelbene

Khalil Damak

Undergraduate Students

Jared Jones

Kiana Cramer

Sam Naser

Grace McClurg

Peer-Reviewed Publications

  • K. Damak, S. Khenissi, O. Nasraoui (2021). A Framework for Unbiased Explainable Pairwise Ranking for Recommendation.  Software Impacts. 100208  .
  • K. Damak, S. Khenissi, O. Nasraoui (2021). Debiased Explainable Pairwise Ranking from Implicit Feedback.  In Proceedings of the 15th ACM Conference on Recommender Systems (ACM RecSys).   321.
  • S. Khenissi, M. Boujelbene, O. Nasraoui. Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System, In Proceedings of the ACM Conference on Recommender Systems (ACM RecSys), 2020
  • M. Badami, O. Nasraoui (2021). PaRIS: Polarization-aware Recommender Interactive System.  In Proceedings of 2021 Workshop on Online Misinformation- and Harm-Aware Recommender Systems (OHARS). 15th ACM Conference on Recommender Systems.  
  • W. Sun and O. Nasraoui (2021). User Polarization Aware Matrix Factorization for Recommendation Systems.  In Proceedings of the 2021 Workshop on Online Misinformation- and Harm-Aware Recommender Systems (OHARS). 15th ACM Conference on Recommender Systems.
  • W. Sun, Grace McClurg, S. Khenissi, O. Nasraoui, and P. Shafto (2022). Debiasing the human-recommender system feedback loop in collaborative filtering..  In preparation for journal submission.
  • Yang, S., C-H., Rank, C., Whritner, J.A., Nasraoui, O., & Shafto, P. (2022). Unifying recommendation and active learning for information filtering and recommender systems. Under Revision  (2022).
  • Shafto, P. & Nasraoui, O. “Human-recommender systems: From benchmark data to benchmark cognitive models” Proceedings of the 10th ACM Conference on Recommender Systems (RecSys ’16) , 2016 , p.127 http://dx.doi.org/10.1145/2959100.2959188
  • Yang, S.C-H., Whritner, J.A., Nasraoui, O. & Shafto, P. “Unifying recommendation and active learning for human-algorithm interactions” Proceedings of the 39th annual conference of the Cognitive Science Society , 2017
  • W. Sun, O. Nasraoui, and P. Shafto “Iterated Algorithmic Bias in the Interactive Machine Learning Process of Information Filtering” in Proceedings of 2018 International Conference on Knowledge Discovery and Information retrieval, Winner of the Best Paper Award. , 2018
  • M. Badami, O. Nasraoui, and P. Shafto “PrCP: Pre-recommendation Counter-Polarization” in Proceedings of 2018 International Conference on Knowledge Discovery and Information retrieval. , 2018
  • M. Badami, O. Nasraoui, W. Sun and P. Shafto “Detecting Polarization in Ratings: An Automated Pipeline and a Preliminary Quantification on Several Benchmark Data Sets” 2017 International Workshop on Big Social Media Data Management, IEEE International Conference on Big Data. , 2017
  • W. Sun, S. Khenissi, O. Nasraoui, and P. Shafto, “Debiasing Collaborative Filtering Recommender Systems” In Augmenting Intelligence with Bias-aware Humans­-in-­the-­Loop (HumBL) workshop, co-located with the Web Conference, San Francisco, CA, 2019. , 2019
  • W. Sun, O. Nasraoui, and P. Shafto, “Evolution and Impact of Bias in Human and Machine Learning Algorithm Interaction” Plos ONE , v.15 , 2020 , p.e0235502

Pre-prints

  • O. Nasraoui, P. Shafto, Human-algorithm Interaction Biases in the Big Data Cycle: A Markov Chain Iterated Learning Framework, arXiv preprint arXiv:1608.07895, 2016.

  • Scott Yang, Chirag Rank, Jake Whritner, Olfa Nasraoui, Patrick Shafto, Unifying recommendation and active learning for information filtering and recommender systems, psycharxiv <https://psyarxiv.com/jqa83>

  • S. Khenissi, O. Nasraoui, Modeling and Counteracting Exposure Bias in Recommender Systems, arXiv preprint arXiv:2001.04832, 2020.

Audio or Video Products.

  • Brief Teaser Video of our Recsys 2020 paper (S. Khenissi, M. Boujelbene, O. Nasraoui. Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System, In Proceedings of the ACM Conference on Recommender Systems (ACM RecSys), 2020) using Simba the puppy as the main star: https://www.youtube.com/watch?v=QO-anl8sC-Q
  • Video of the presentation by Pat Shafto at RecSys 2016 of the paper by Pat Shafto and Olfa Nasraoui “Human Recommender Systems: From Benchmark Data to Benchmark Cognitive Models”: https://youtu.be/2XiDQDHIhsI
  • Video of Sami Khenissi’s presentation of our Recsys 2020 paper (S. Khenissi, M. Boujelbene, O. Nasraoui. Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System, In Proceedings of the ACM Conference on Recommender Systems (ACM RecSys), 2020): https://www.youtube.com/watch?v=QO-anl8sC-Q

Software

Software

Software

Brief Research Explainer to a General Audience.

  • In the paper/website below, we provide a brief Explainer to a General Audience of our RecSys2021 paper:
  • Damak, Khalil; Khenissi, Sami; and Nasraoui, Olfa. Promoting explainability and unbiasedness in ranking-based recommendation in Kudos, Association for Computing Machinery (ACM), Sep. 2021.
  • https://www.growkudos.com/publications/10.1145%252F3460231.3474274/reader

Other Publications

  • Damak, Khalil; Khenissi, Sami; and Nasraoui, Olfa. (2021). Promoting Explainability and Unbiasedness in Ranking-Based Recommendation.. 

PhD Dissertations

  • Sami Khenissi. Modeling And Debiasing Feedback Loops In Collaborative Filtering Recommender Systems. (2022).  University of Louisville.
  • Khalil Damak. New Accurate, Explainable, and Unbiased Machine Learning Models for Recommendation with Implicit Feedback. (2022).  University of Louisville.
  • Mahsa Badami. Peeking into the Other Half of The Glass: Handling Polarization in Recommender Systems. (2017)
  • Wenlong Sun. Studying and Handling Iterated Algorithmic Biases in Human and Machine Learning Interaction. (2019).  University of Louisville.
  • Mariem Boujelbene. “Multiple Bias Correction in Collaborative Filtering Recommender Systems. (2022).  University of Louisville.
  • Aneseh Alvanpour. Beyond Accuracy in Machine Learning. (2022).  University of Louisville.

Masters Thesis

  • Sami Khenissi. Modeling and Counteracting Exposure Bias in Recommender Systems (MS Thesis). (2019)

Websites or Other Internet Sites