Participatory Personalization in Classification

Machine learning models are often personalized with information that is protected, sensitive, self-reported, or costly to acquire. These models use information about people but do not facilitate nor inform their consent. Individuals cannot opt out of reporting personal information to a model, nor tell if they benefit from personalization in the first place. We introduce a family of classification models, called participatory systems, that let individuals opt into personalization at prediction time. We present a model-agnostic algorithm to learn participatory systems for personalization with categorical group attributes. We conduct a comprehensive empirical study of participatory systems in clinical prediction tasks, benchmarking them with common approaches for personalization and imputation. Our results demonstrate that participatory systems can facilitate and inform consent while improving performance and data use across all groups who report personal data.

Paper

References (68)

Scroll for more · 38 remaining

Similar papers

Peer review

Reviewer ekFJ6/10 · confidence 4/52023-06-24

Summary

Machine learning models often use personal and sensitive information without facilitating or informing individuals' consent. This paper introduces a family of classification models called participatory systems, which allow individuals to opt into personalization at prediction time. The study demonstrates that these systems can improve performance and data use while facilitating and informing consent for those who report personal data.

Strengths

The paper studied a practical problem of compromising ML model performance and personal information consenting. The idea of constructing the reporting interface tree and pruning based on the models' performance along the path is reasonable and sound. The experiments showed promising results.

Weaknesses

The proposed method will become complicated when the number of categorical attributes increase.

Questions

1. Are all personal info available during training time? This study reminds me of the research on missing data. The personal info consent will also lead to missing data for the later rounds of retraining. It is actually a Missing-not-at-radom problem where people choose to opt out of personalization with some concerns, which definitely are not at random.

Rating

6: Weak Accept: Technically solid, moderate-to-high impact paper, with no major concerns with respect to evaluation, resources, reproducibility, ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

3 good

Presentation

3 good

Contribution

3 good

Limitations

The authors have discussed the limitation that data collection may influence the later rounds of retraining.

Reviewer SNn47/10 · confidence 4/52023-06-29

Summary

This paper presents a new approach for operating machine learning systems in settings where results are personalized in some fashion to individuals or group (i.e., some categorical feature that “personalizes” predictions). The goal of this new approach is to make the process of personalizing ML results more participatory for data subjects / data contributors, moving beyond the existing practices that are lacking when it comes to facilitating informed consent of data contributors. Specifically, the authors propose that by adopting the presented “participatory systems” approach, individuals who are subject to ML predictions can make decisions about sharing personal information at prediction time and better understand how information sharing impacts ML outputs. Further, the paper proposes an approach to help individuals navigate the opt in / opt out tradeoff. The basic idea underlying the paper is to limit data use where personalization does not improve performance. The authors use clinical prediction models as a running use-case. Motivating examples include personalization based on features that are protected groups, self-reported by patients, or costly to acquire (e.g., requires doctor to label). The paper first provides some relevant definitions that are mathematical in nature. An algorithm for “learning participatory systems”, i.e., selecting from a pool of personalized models and pruning away models that violate incentive compatibility, is presented. Then, the paper presents some experiments using the clinical prediction setting.

Strengths

Overall, I found this paper to paint a compelling vision of an alternate paradigm for ML. The overall quality was high, and I believe this paper can make a large contribution to the ML research community. The originality is high. I think this draft does a great job describing problems with the status quo around personalized modeling. While the work certainly draws on existing scholarship and practical work in privacy, human-computer interaction, and FAccT, the specific formalization presented here is original. See below for several lines of work that may be particularly relevant (though I understand if the authors see this as out of scope). The paper makes a compelling case for the importance of this line of work. Motivation was strong in general and could attract more interest in the broad class of participatory systems. Overall clarity of the paper is high. The primary task being studied here is described well. On the whole, assumptions are stated explicitly and the claims in the paper are precisely stated. The presentation of ideas and results is also strong. Figure 2 and Table 1 are particularly helpful.

Weaknesses

Overall, the current draft makes a strong contribution. A high-level weakness of the paper is that this is very much a user-centric / human-centric paper (the core idea is help data subjects) but no evaluation with actual users/subjects is included. Nonetheless, I think this is still a reasonable starting point for advancing this line of work. Several other areas for potential improvement: - The proposed approach / framework could be easier to follow with a summary of assumptions about the behavior of relevant agents, especially ML operators. I understood the basic assumption about groups wanting to maximizing accuracy for their group, but I think readers will want more on how different ML operator incentives matter here. - There’s a lot of related work in human-computer interaction, privacy, and data economics that could be discussed more heavily. This may be out of scope, however. - The paper could more explicit about how the resulting training datasets different when using this approach. The intended user experience for data subjects is very clear (i.e. I understand how a patient might interact with participatory systems), but it wasn’t totally clear if ML operators who use this approach will actually have substantially different looking training sets. Will ML operators eventually lose the ability to evaluate models for certain groups if those groups choose not to participate? In other words, what does this look for the perspective of a data engineer? - Relatedly, while the motivation is strong, I’d like to see a bit more discussion of how ML operators would adopt this approach. It seems the actual interface design could be a major barrier to use.

Questions

The authors may wish to engage with some recent work on data minimization. - Biega, A.J., Potash, P., Daumé, H., Diaz, F. and Finck, M., 2020, July. Operationalizing the legal principle of data minimization for personalization. In Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval (pp. 399-408). - Shanmugam, D., Diaz, F., Shabanian, S., Finck, M. and Biega, A., 2022, June. Learning to limit data collection via scaling laws: A computational interpretation for the legal principle of data minimization. In 2022 ACM Conference on Fairness, Accountability, and Transparency (pp. 839-849). Similarly, some recent work on collective action by data contributors may be useful for this paper (or future work, as this discussion may be out of scope). This work has great potential to scaffold and support data-related action by groups, perhaps to enable bargaining between data creators and data users. - Vincent, N., Li, H., Tilly, N., Chancellor, S. and Hecht, B., 2021, March. Data leverage: A framework for empowering the public in its relationship with technology companies. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (pp. 215-227). - Das, S., Edwards, W.K., Kennedy-Mayo, D., Swire, P. and Wu, Y., 2021. Privacy for the people? exploring collective action as a mechanism to shift power to consumers in end-user privacy. IEEE Security & Privacy, 19(5), pp.66-70. In the Introduction, a distinction between online platforms (that do use informed consent) and personalized models (that do not) is made. I think this distinction is not quite as clear as the current draft implies; online platforms and models are highly intertwined in many contexts. This might be downstream of the focus on medical personalization, but I think there are many other domains which much larger “grey area” between platforms data collection and data collection for modeling. Minor - Duplicate refs: [4] and [5] - Line 57 missing bracket

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

3 good

Presentation

4 excellent

Contribution

4 excellent

Limitations

The work is heavily focused on societal impacts of ML. On this front, I think the paper is very strong. On the methods limitations front, it might help readers to talk a bit more about the limitations of evaluating participatory systems with offline evaluation techniques, as opposed to user studies.

Reviewer LFwB7/10 · confidence 3/52023-07-06

Summary

In this paper, the authors propose an innovative solution, dubbed 'participatory systems', to address ethical and pragmatic issues arising from the personalization process in machine learning models. This work timely addresses an escalating concern in machine learning, specifically the utilization of sensitive or protected personal data without an individual's explicit consent and without visible individual benefits.

Strengths

The authors offer a comprehensive examination of the problem, substantiated with compelling real-world examples, that underlines the urgent need for a solution like the one proposed. They argue, persuasively, that traditional personalization can occasionally be counterproductive, either failing to improve or even worsening outcomes for certain groups. The authors convincingly contest the notion that models should automatically presume the availability of all data used during training at the time of prediction.

Weaknesses

However, despite the clarity of the explanation and visualization, it remains unclear why exactly the proposed framework functions as it does (mechanism). In addition, the authors should consider clearly outlining the inherent assumptions underpinning their model. For instance, they could explore the potential impacts of widespread simultaneous opt-outs due to specific events, such as a controversial news story affecting a cohort of users. They should consider scenarios where a majority of users opt out during a particular period or in response to a certain question. This would provide practitioners with crucial guidance regarding the applicability and potential limitations of the proposed system. The authors could enhance the robustness of the paper by more thoroughly discussing potential challenges and limitations of implementing participatory systems in real-world settings. For instance, it is not clear how these systems would manage the variability in individuals' capacity or willingness to share personal data or how they would account for potential disparities in users' technological literacy. Furthermore, while the authors propose that the participatory system can curtail unnecessary data collection, there is no clear discussion on how the system would manage potential data sparsity resulting from this reduction. Another improvement could be a more comprehensive exploration of the applicability of participatory systems to other types of personalization models or data types, such as continuous attributes or unstructured data. A thorough discussion of this aspect would provide additional value. The paper could also benefit from an in-depth discussion on the practical hurdles and considerations in implementing participatory systems. Factors such as scalability, user interface design, and potential biases in opt-in/opt-out decisions should be considered to provide a more holistic understanding of the system's feasibility.

Questions

See above

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

3: You are fairly confident in your assessment. It is possible that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.

Soundness

3 good

Presentation

3 good

Contribution

3 good

Limitations

See above

Reviewer tDFi3/10 · confidence 4/52023-07-08

Summary

The paper introduces Participatory Personalization in Classification (PPC), a new family of classification models that enables individuals to opt into personalization at prediction time. It argues that existing personalization models lack consent and do not allow individuals to opt out of sharing personal information or assess the benefits of personalization. PPC models address these concerns by allowing individuals to choose personalization, leading to improved performance and informed consent. The paper presents a theoretical framework for PPC and evaluates its effectiveness in various domains, such as healthcare and recommender systems. The authors propose a model-agnostic algorithm for learning PPC models with categorical group attributes. They conduct an extensive empirical study on clinical prediction tasks, comparing PPC models to common personalization and imputation approaches.

Strengths

1. The paper presents a novel approach to personalization that allows individuals to opt into personalization at prediction time, leading to improved performance and informed consent. 2. The authors provide a robust theoretical framework for Participatory Personalization in Classification (PPC) and demonstrate its effectiveness in various real-world applications, including healthcare and recommender systems. 3. The paper offers a comprehensive review of related work, providing valuable insights into the limitations of existing personalization approaches.

Weaknesses

1. A detailed comparison of PPC with other approaches to personalization, such as collaborative filtering or content-based filtering, is lacking, which could provide a more comprehensive understanding of PPC's strengths and limitations in relation to existing methods. 2. The computational complexity of PPC and its scalability to large datasets are not extensively analyzed, leaving uncertainty about the feasibility of applying PPC in scenarios with substantial data volumes. 3. The limitations of the real-world applications of PPC, including data availability and quality, are not extensively discussed, which could provide insights into the practical challenges of implementing PPC in different domains. 4. A detailed analysis of the limitations of the theoretical framework of PPC, including assumptions and model constraints, is missing, hindering a comprehensive understanding of the model's applicability and potential shortcomings.

Questions

1. Have you considered the potential for PPC models to exacerbate existing biases or discrimination in the data? If so, how have you addressed these concerns in your empirical study? 2. How does the computational complexity of the PPC algorithm compare to other approaches to personalization, and how scalable is it to large datasets? 3. Have you considered the potential limitations of the real-world applications of PPC, such as the availability and quality of data, and how might these limitations affect the effectiveness of the model in practice?

Rating

3: Reject: For instance, a paper with technical flaws, weak evaluation, inadequate reproducibility and incompletely addressed ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

3 good

Presentation

2 fair

Contribution

2 fair

Limitations

The ethical implications of PPC, such as potential discrimination or bias, are not thoroughly evaluated, warranting further investigation.

Reviewer SNn42023-08-11

Helpful rebuttal

Thanks to the authors for this rebuttal. This is all very helpful additional information, and I think the planned minor revisions and discussion points raised here all (1) seem very feasible and (2) help to directly address some of the opportunities for improvement identified in reviews. I would argue for accepting.

Reviewer ekFJ2023-08-19

I've read the authors' reply and other reviewers's comments. I will keep my rating.

Reviewer LFwB2023-08-20

I thank the authors for their helpful rebuttals, as well as their clarifications and references. I would argue for acceptance of this paper.

Area Chair vMsm2023-08-21

I would like to thank the authors for their thorough clarification and response to some of the reviewers major concerns. AC

Program Chairsdecision2023-09-21

Decision

Accept (spotlight)

© 2026 NYSGPT2525 LLC