pith. sign in

arxiv: 1907.07253 · v1 · pith:EBT7FZXHnew · submitted 2019-07-16 · 💻 cs.SI · cs.IR

Fairness and Diversity in the Recommendation and Ranking of Participatory Media Content

Muskaan , Mehak Preet Dhaliwal , Aaditeshwar Seth This is my paper

Pith reviewed 2026-05-24 20:18 UTC · model grok-4.3

classification 💻 cs.SI cs.IR
keywords fairnessdiversityrecommendation systemsparticipatory mediavoice-based platformsuser-generated contentranking algorithms
0
0 comments X

The pith

A model ranks participatory media content to give fair and diverse exposure to different viewpoints on a topic.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a recommendation and ranking model for user-generated audio content on a voice-based platform serving low-income and less-literate users in rural central India. The model aims to ensure that multiple viewpoints receive exposure rather than allowing any single perspective to dominate the ranked list played during phone calls. Evaluation compares the model's performance on fairness and diversity metrics against the manual editorial ranking currently used, drawing on platform call logs. The approach is presented as adaptable to other participatory media systems that face similar volume and viewpoint-balance challenges.

Core claim

We address the problem of recommending and ranking this content such that different viewpoints about a topic get exposure in a fair and diverse manner. We build our model in the context of a voice-based participatory media platform running in rural central India, for low-income and less-literate communities, that plays audio messages in a ranked list to users over a phone call and allows them to contribute their own messages. Our models are generic and can be adapted and applied to other participatory media platforms as well.

What carries the argument

A ranking model that incorporates fairness and diversity constraints when ordering user-generated audio messages for phone-call playback.

If this is right

  • The same ranking approach can be used on other participatory media platforms without requiring platform-specific redesign.
  • Platforms can reduce dependence on human editors while still tracking measurable fairness and diversity outcomes.
  • Users receive ranked lists that systematically surface multiple viewpoints rather than a narrow selection.
  • New content contributions can be inserted into the ranking pipeline while preserving exposure balance across topics.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the model generalizes, similar fairness constraints could be tested on text-based social platforms to limit viewpoint concentration.
  • The evaluation method using call logs suggests a path for platforms to audit ranking effects without direct user surveys.
  • Extending the model to track viewpoint evolution over time could reveal whether repeated exposure shifts user contributions.

Load-bearing premise

Call-log data from the platform can accurately measure whether the model achieves better fairness and diversity than manual editorial processes.

What would settle it

Running the model on a new set of call logs and finding that its viewpoint-exposure metrics show no improvement or are worse than those produced by the manual ranking process.

Figures

Figures reproduced from arXiv: 1907.07253 by Aaditeshwar Seth, Mehak Preet Dhaliwal, Muskaan.

Figure 1
Figure 1. Figure 1: Content Pipeline: Stages of processing undergone by new items which are recorded by users [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Identification of users for evaluation of the algorithms [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of users across ranks in a list and differ [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of models. Figures a and b show combined results for users of both the clusters [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cumulative distribution of exposure achieved by the items for the ratings 3,4 and 5 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
read the original abstract

Online participatory media platforms that enable one-to-many communication among users, see a significant amount of user generated content and consequently face a problem of being able to recommend a subset of this content to its users. We address the problem of recommending and ranking this content such that different viewpoints about a topic get exposure in a fair and diverse manner. We build our model in the context of a voice-based participatory media platform running in rural central India, for low-income and less-literate communities, that plays audio messages in a ranked list to users over a phone call and allows them to contribute their own messages. In this paper, we describe our model and evaluate it using call-logs from the platform, to compare the fairness and diversity performance of our model with the manual editorial processes currently being followed. Our models are generic and can be adapted and applied to other participatory media platforms as well.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a model for recommending and ranking user-generated audio content on a voice-based participatory media platform operating in rural central India. The goal is to ensure fair and diverse exposure of different viewpoints on topics. The model is evaluated using call-logs from the platform, with claims of improved fairness and diversity performance relative to the manual editorial processes currently in use. The approach is presented as generic and adaptable to other participatory media platforms.

Significance. If the evaluation is valid, the work could offer a practical algorithmic approach to viewpoint diversity in recommendation systems for low-literacy, voice-based platforms, addressing a real-world curation challenge. The use of real platform data for comparison is a positive aspect, though the strength of the contribution hinges on whether the metrics truly capture exposure equity.

major comments (2)
  1. [Evaluation] Evaluation section: The central claim that the proposed model improves fairness and diversity over manual curation rests on metrics computed from call-logs. However, the manuscript provides no explicit description of how individual messages are labeled with viewpoints, how selection frequency or listening duration is mapped to exposure, or any validation that these logs serve as a stable proxy for equitable viewpoint perception across users. This mapping is load-bearing for the empirical comparison.
  2. [Model] Model description (likely §3): While the abstract states that the model ensures different viewpoints get exposure, the text does not detail the specific fairness or diversity objective functions, constraints, or optimization procedure used to rank content. Without these, it is not possible to assess whether the approach is parameter-free or how it differs substantively from standard ranking methods.
minor comments (2)
  1. [Abstract] The abstract lacks any mention of the concrete metrics, datasets, or quantitative results from the call-log evaluation, making it difficult for readers to gauge the strength of the claims at first reading.
  2. Notation for fairness and diversity quantities should be defined consistently when first introduced, and any tables comparing model vs. manual performance should include confidence intervals or statistical significance tests.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on our manuscript. We address each major point below and will make revisions to improve clarity and completeness where the feedback identifies gaps in the current text.

read point-by-point responses

  1. Referee: [Evaluation] Evaluation section: The central claim that the proposed model improves fairness and diversity over manual curation rests on metrics computed from call-logs. However, the manuscript provides no explicit description of how individual messages are labeled with viewpoints, how selection frequency or listening duration is mapped to exposure, or any validation that these logs serve as a stable proxy for equitable viewpoint perception across users. This mapping is load-bearing for the empirical comparison.

    Authors: We agree that the evaluation section would benefit from greater explicitness. The original manuscript describes the call-log data and the comparison to manual curation but does not spell out the viewpoint labeling procedure (performed by platform moderators according to topic and expressed stance) or the exact mapping from logs to exposure (cumulative selections multiplied by average listening duration per message). We will revise the evaluation section to include these details, along with a brief discussion of the logs as a behavioral proxy and its limitations. This constitutes a clarification rather than a change to the underlying analysis. revision: yes

  2. Referee: [Model] Model description (likely §3): While the abstract states that the model ensures different viewpoints get exposure, the text does not detail the specific fairness or diversity objective functions, constraints, or optimization procedure used to rank content. Without these, it is not possible to assess whether the approach is parameter-free or how it differs substantively from standard ranking methods.

    Authors: We acknowledge that §3 presents the model at a conceptual level without the full mathematical specification. The approach combines a relevance term with fairness (L1 deviation from uniform viewpoint distribution) and diversity (coverage of viewpoint pairs) terms inside a constrained ranking formulation solved by a greedy selection procedure with tunable trade-off parameters. We will expand the model section with the explicit objective, constraints, and optimization steps so that readers can evaluate its relation to standard methods. revision: yes

Circularity Check

0 steps flagged

No significant circularity in model derivation or evaluation.

full rationale

The paper proposes a recommendation model for fairness and diversity of viewpoints in participatory media and evaluates performance via comparison to manual curation using independent call-log data from the platform. No equations, self-definitional steps, fitted parameters renamed as predictions, or load-bearing self-citations are present in the provided text. The central claims rest on external data and direct comparison rather than reducing to inputs by construction, so the derivation chain is self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no information on parameters, axioms, or entities is provided.

pith-pipeline@v0.9.0 · 5685 in / 890 out tokens · 19898 ms · 2026-05-24T20:18:45.559933+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages · 3 internal anchors

  1. [1]

    Himan Abdollahpouri, Robin Burke, and Bamshad Mobasher. 2017. Control- ling popularity bias in learning-to-rank recommendation. In Proceedings of the Eleventh ACM Conference on Recommender Systems. ACM, 42–46

    work page 2017

  2. [2]

    Adamic and Natalie Glance

    Lada A. Adamic and Natalie Glance. 2005. The Political Blogosphere and the 2004 U.S. Election: Divided They Blog. In Proceedings of the 3rd International Workshop on Link Discovery (LinkKDD ’05). ACM, New York, NY , USA, 36–43. https://doi.org/10.1145/1134271.1134277

    work page doi:10.1145/1134271.1134277 2005

  3. [3]

    Chinmayi Arun. 2019. On WhatsApp, Rumours, and Lynchings. Economic & Political Weekly54, 6 (2019), 31

    work page 2019

  4. [4]

    Eytan Bakshy, Solomon Messing, and Lada A Adamic. 2015. Exposure to ide- ologically diverse news and opinion on Facebook. Science 348, 6239 (2015), 1130–1132

    work page 2015

  5. [5]

    Asia J Biega, Krishna P Gummadi, and Gerhard Weikum. 2018. Equity of attention: Amortizing individual fairness in rankings. arXiv preprint arXiv:1805.01788 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  6. [6]

    Ranking with Fairness Constraints

    L. Elisa Celis, Damian Straszak, and Nisheeth K. Vishnoi. 2017. Ranking with Fairness Constraints. arXiv e-prints, Article arXiv:1704.06840 (Apr 2017), arXiv:1704.06840 pages. arXiv:cs.DS/1704.06840

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  7. [7]

    Dipanjan Chakraborty, Akshay Gupta, and Aaditeshwar Seth. 2019. Experiences from a mobile-based behaviour change campaign on maternal and child nutrition in rural India. InProceedings of the Tenth International Conference on Information and Communication Technologies and Development. ACM, 20

    work page 2019

  8. [8]

    Nitesh V Chawla, Kevin W Bowyer, Lawrence O Hall, and W Philip Kegelmeyer

    work page

  9. [9]

    Journal of artificial intelligence research 16 (2002), 321–357

    SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research 16 (2002), 321–357

    work page 2002

  10. [10]

    Michael D Conover, Jacob Ratkiewicz, Matthew Francisco, Bruno Gonçalves, Filippo Menczer, and Alessandro Flammini. 2011. Political polarization on twitter. In Fifth international AAAI conference on weblogs and social media

    work page 2011

  11. [11]

    Daniel A Crane and I All. 2012. Search Neutrality as an antitrust principle. 19 Geo. Mason L. Rev 1199 (2012), 2011–2012

    work page 2012

  12. [12]

    Amit Datta, Michael Carl Tschantz, and Anupam Datta. 2015. Automated exper- iments on ad privacy settings. Proceedings on privacy enhancing technologies 2015, 1 (2015), 92–112

    work page 2015

  13. [13]

    Larry Diamond. 2015. Liberation technology 1. In In Search of Democracy . Routledge, 132–146

    work page 2015

  14. [14]

    Robert Epstein and Ronald E Robertson. 2015. The search engine manipulation effect (SEME) and its possible impact on the outcomes of elections. Proceedings of the National Academy of Sciences 112, 33 (2015), E4512–E4521

    work page 2015

  15. [15]

    Ayman Farahat and Michael C Bailey. 2012. How effective is targeted advertising?. In Proceedings of the 21st international conference on World Wide Web. ACM, 111–120

    work page 2012

  16. [16]

    Pratik Gajane and Mykola Pechenizkiy. 2017. On formalizing fairness in predic- tion with machine learning. arXiv preprint arXiv:1710.03184 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  17. [17]

    Catherine Grevet, Loren G Terveen, and Eric Gilbert. 2014. Managing politi- cal differences in social media. In Proceedings of the 17th ACM conference on Computer supported cooperative work & social computing. ACM, 1400–1408

    work page 2014

  18. [18]

    James Grimmelmann. 2010. Some skepticism about search neutrality. The next digital decade: Essays on the future of the Internet (2010), 435

    work page 2010

  19. [19]

    Andrew Guess, Brendan Nyhan, and Jason Reifler. 2018. Selective exposure to misinformation: Evidence from the consumption of fake news during the 2016 US presidential campaign. European Research Council 9 (2018)

    work page 2018

  20. [20]

    Jurgen Habermas, Jürgen Habermas, and Thomas Mccarthy. 1991. The structural transformation of the public sphere: An inquiry into a category of bourgeois society. MIT press

    work page 1991

  21. [21]

    Moritz Hardt, Eric Price, Nati Srebro, et al . 2016. Equality of opportunity in supervised learning. In Advances in neural information processing systems. 3315– 3323

    work page 2016

  22. [22]

    Jie Lu, Dianshuang Wu, Mingsong Mao, Wei Wang, and Guangquan Zhang. 2015. Recommender system application developments: a survey. Decision Support Systems 74 (2015), 12–32

    work page 2015

  23. [23]

    Aaditeshwar Seth Mehak, Muskaan. [n. d.].Supplementary material: Fairness and Diversity in the Recommendation and Ranking of Participatory Media Content. available at: https://bit.ly/30oY4Wr. Fairness and Diversity in Recommendation Systems Intelligent Information Feed ’19, August 05, 2019, Anchorage, Alaska, USA

    work page 2019

  24. [24]

    Aparna Moitra, Vishnupriya Das, Gram Vaani, Archna Kumar, and Aaditeshwar Seth. 2016. Design Lessons from Creating a Mobile-based Community Media Platform in Rural India. In Proceedings of the Eighth International Conference on Information and Communication Technologies and Development. ACM, 14

    work page 2016

  25. [25]

    Preeti Mudliar, Jonathan Donner, and William Thies. 2012. Emergent Practices Around CGNet Swara, V oice Forum for Citizen Journalism in Rural India. In Proceedings of the Fifth International Conference on Information and Communi- cation Technologies and Development (ICTD ’12). ACM, New York, NY , USA, 159–168. https://doi.org/10.1145/2160673.2160695

    work page doi:10.1145/2160673.2160695 2012

  26. [26]

    Cathy O’Neil. 2017. Weapons of math destruction: How big data increases inequality and threatens democracy. Broadway Books

    work page 2017

  27. [27]

    Neil Patel, Deepti Chittamuru, Anupam Jain, Paresh Dave, and Tapan S. Parikh

    work page

  28. [28]

    In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10)

    Avaaj Otalo: A Field Study of an Interactive V oice Forum for Small Farmers in Rural India. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10). ACM, New York, NY , USA, 733–742. https://doi.org/10.1145/1753326.1753434

    work page doi:10.1145/1753326.1753434

  29. [29]

    Stephen A Rhoades. 1993. The herfindahl-hirschman index. Fed. Res. Bull. 79 (1993), 188

    work page 1993

  30. [30]

    Dietram A Scheufele and David Tewksbury. 2006. Framing, agenda setting, and priming: The evolution of three media effects models. Journal of communication 57, 1 (2006), 9–20

    work page 2006

  31. [31]

    Aaditeshwar Seth. 2019. Ensuring Responsible Outcomes from Technology, COMSNETS. (2019)

    work page 2019

  32. [32]

    Aaditeshwar Seth and Jie Zhang. 2008. A Social Network Based Approach to Personalized Recommendation of Participatory Media Content.. In ICWSM

    work page 2008

  33. [33]

    Aaditeshwar Seth, Jie Zhang, and Robin Cohen. 2015. A personalized credibility model for recommending messages in social participatory media environments. World Wide Web18, 1 (2015), 111–137

    work page 2015

  34. [34]

    Ashudeep Singh and Thorsten Joachims. 2018. Fairness of exposure in rankings. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM, 2219–2228

    work page 2018

  35. [35]

    Samanth Subramanian. 2017. Inside the macedonian fake-news complex. Wired magazine 15 (2017)

    work page 2017

  36. [36]

    Soroush V osoughi, Deb Roy, and Sinan Aral. 2018. The spread of true and false news online. Science 359, 6380 (2018), 1146–1151

    work page 2018

  37. [37]

    Ke Yang and Julia Stoyanovich. 2017. Measuring fairness in ranked outputs. In Proceedings of the 29th International Conference on Scientific and Statistical Database Management. ACM, 22

    work page 2017

  38. [38]

    Meike Zehlike, Francesco Bonchi, Carlos Castillo, Sara Hajian, Mohamed Mega- hed, and Ricardo Baeza-Yates. 2017. Fa* ir: A fair top-k ranking algorithm. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management. ACM, 1569–1578

    work page 2017