Designing social media algorithms for civic resilience instead of engagement. Explores algorithmic objectives that prioritize informational quality and social cohesion. Reviews academic proposals and prototypes, and outlines implementation hurdles.
Executive summary
The literature contains a small but clear set of academic proposals that move social media ranking away from pure engagement and toward civic or social outcomes, including democratic discourse, offline civic participation, collective well-being, content quality, and social cohesion[1][2][3][4].
Evidence is uneven: the strongest items are conceptual proposals, while the most concrete tests are audits, field experiments, and a participatory prototype. Across those tests, changes in ranking or exposure can alter divisive content consumption and even affective polarization, but the work also shows important trade-offs and unresolved deployment questions[5][6][7].
Academic proposals for civic-resilience-oriented ranking
Several papers sketch alternative objectives for recommender systems. One proposes "alternative content recommendation algorithms" grounded in healthy civic discourse, the EU Digital Services Act, and democratic values, while explicitly discussing the trade-off between intervention and freedom of expression[8]. Another gives a framework for algorithms that may promote offline civic participation through attitudes, motivation, and social capital[9].
- A Collective Well-Being aware Recommender System (CWB-RS) is proposed inside an educational social media companion, with the goal of balancing external platform recommendations against a community-level well-being objective rather than engagement[10][11].
- The CWB-RS proposal says the system should help users access a more diverse set of perspectives and sources, provide opposite but not confrontational perspectives, reduce racist content by substituting low-harm feeds, and steer learners toward authoritative resources rather than extreme individual opinions[12][13][14].
- A separate normative paper proposes a "right to constructive optimization" for recommender systems under the Digital Services Act, framing the objective in terms of self-determination, self-development, inclusivity, political equality, reasonableness, and publicity[15].
- Adjacent quality and veracity ideas also appear in the sources: a Doximity audit emphasizes assessing content quality, and a fake-news review argues that crowdsourced veracity ratings can improve social media ranking algorithms[16][17].
- These proposals are mostly conceptual rather than deployed systems or completed field tests[18][19].
Prototypes and empirical tests
The clearest tested evidence in the supplied sources comes from ranking audits and live experiments, not from fully deployed civic-optimization platforms. These studies matter because they show that changing ranking rules can measurably change what users see and how they respond politically[20][21].
- A preregistered field experiment on X reranked feeds in real time with an LLM to increase or decrease exposure to antidemocratic attitudes and partisan animosity, and found that changing exposure shifted out-party partisan animosity by more than 2 points on a 100-point feeling thermometer[22].
- A Twitter audit compared the engagement-based timeline with a reverse-chronological baseline and a stated-preference timeline built from users' survey responses. It found that engagement-based ranking amplified more partisan, emotionally negative, and out-group hostile content, while the stated-preference ranking reduced negativity and divisiveness but could increase exposure to in-group content[23].
- A middle-school participatory design workshop produced the Social Sketch prototype, which experimented with removing likes and comments, using view counts instead, adding tags and filters, and exploring anonymity and personalization as ways to support social wellbeing[24].
- For informational quality, the Doximity paper and the fake-news review are best read as adjacent evidence that quality and veracity can be treated as ranking objectives, but they are not direct field tests of civic-resilience ranking[25][26].
Implementation hurdles and evaluation gaps
The main hurdle is defining a valid objective function. Across the sources, the candidate targets differ: civic participation, democratic discourse, collective well-being, quality, veracity, social cohesion, and social confidence. The literature does not yet show a single validated metric that can reliably unify those goals for deployment[27][28][29][30].
A second hurdle is trade-off management. The Twitter audit warns that optimizing stated preferences could increase in-group bias, the societal-good paper highlights tension with freedom of expression, and the audit paper says more research is needed before real-world deployment because of technical and platform-incentive challenges[31][32].
A third hurdle is evaluation. The strongest evidence today comes from narrow experiments, audits, and prototypes, which means the field still lacks broad proof that civic-optimized ranking improves downstream civic resilience at scale[33][34][35].
Compact evidence table
| Item | Study type | Main contribution | Limitation or status |
|---|---|---|---|
| Designing social media content recommendation algorithms for societal good[36] | Perspective / PubMed record[37] | Proposes alternative recommendation algorithms rooted in civic discourse, democratic values, and the EU Digital Services Act[38] | Conceptual proposal, not a deployed system[39] |
| How Social Media Algorithms Shape Offline Civic Participation[40] | Theoretical framework / review article[41] | Frames how algorithms may influence offline civic participation through attitudes, motivation, and social capital[42] | Framework paper rather than an implementation test[43] |
| Challenging social media threats using collective well-being-aware recommendation algorithms and an educational virtual companion[44] | Proposal paper[45] | Introduces a CWB-RS and virtual companion that balance platform recommendations around collective well-being[46] | Conceptual and educational, not a completed field deployment[47] |
| A Right to Constructive Optimization[48] | Normative legal and policy paper[49] | Defines a public-interest benchmark for recommender systems under the Digital Services Act[50] | Normative framework, not an empirical test[51] |
| Engagement, user satisfaction, and the amplification of divisive content on social media[52] | Preregistered audit / counterfactual ranking comparison[53] | Compares engagement, reverse-chronological, and stated-preference feeds and shows different divisiveness effects[54] | Authors note concerns about in-group bias, technical deployment, and platform incentives[55] |
| Reranking partisan animosity in algorithmic social media feeds alters affective polarization[56] | Field experiment[57] | Shows that live feed reranking can shift partisan animosity[58] | Strong causal evidence, but limited to a specific platform and outcome window[59] |
| Prototyping for Social Wellbeing with Early Social Media Users[60] | Participatory design prototype[61] | Tests design changes such as removing likes and comments, using view counts, and adding filters or anonymity[62] | Prototype stage, not a population-scale ranking system[63] |
Conclusion
Overall, the supplied literature suggests that civic-resilience ranking is best understood as a portfolio of alternative objectives, not a single replacement for engagement. The field already has credible conceptual proposals and a few meaningful tests, but it still needs validated civic metrics, clearer treatment of free-expression and bias trade-offs, and more evidence on whether these designs work beyond audits and prototypes[64][65][66][67].
Get more accurate answers with Super Pandi, upload files, personalized discovery feed, save searches and contribute to the PandiPedia.
Let's look at alternatives:
- Modify the query.
- Start a new thread.
- Remove sources (if manually added).
Curated by Joan