Algorithmic Attention and Content Creation on Social Media Platforms

arXiv:2609.15573 · econ.TH · Submitted 2026-09-14 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We're looking at the paper: "Algorithmic Attention and Content Creation on Social Media Platforms".

Warren: Here's the gist of the paper. We study revenue-maximizing attention allocation on an ad-funded social media platform governed by recommendation algorithms The optimal design uses attention subsidies discriminatorily.

Ruth: The Core Trade-off in Mechanism Design The central economic question addressed is the trade-off between harvesting attention through advertising and cultivating it by strengthening production incentives.

Warren: The platform operates as a two-sided marketplace that hosts content across multiple horizontally differentiated categories and serves viewers and creators <ref:2609.15573#pg2> Designers must balance monetization incentives against the need to provide creators with sufficient exposure to sustain the production of engaging content <ref:2609.15573#pg3>.

Ruth: The framework characterizes the advertisement-revenue-maximizing allocation of attention <ref:2609.15573#pg5>.

Paper summary: Ruth: Thinking about the title "Algorithmic Attention and Content Creation on Social Media Platforms" by Yi Chen Fei Li Marcel Preuss, what’s the big picture they’re pointing toward?

Warren: They are looking at how the way attention is allocated isn't just a technical detail; it directly affects what content gets made and what users end up seeing.

Iris: The paper shows that the optimal mechanism involves targeting distortions based on creator ability, excluding low-ability creators while subsidizing high-ability creators through exposure or monetary payments.

Ruth: It’s about how you manage the trade-off between harvesting attention through advertising and cultivating it by strengthening production incentives.

Warren: The comparative static they run suggests that if you allow direct monetary payments to creators, the optimal mechanism might use both exposure and payment together, which changes things compared to just focusing on one or the other.

Iris: The implementation side of the paper shows this mapping where observed quality is turned into a promised total attention through an attention allocation rule that depends on the content quality.

Ruth: And what they suggest there’s a regulatory trade-off: limiting how much user characteristics are used in recommendations might help participating users by leaving them information rents, but it could also reduce diversity in the platform's content ecosystem.

Warren: That's a complex policy implication because you have to weigh individual welfare against the overall variety of content available to everyone.

Iris: The discussion points toward a stark implication regarding category survival and reach information asymmetry, suggesting how regulations like the DMA might affect that attention-allocation trade-off.

Ruth: It really forces us to think about how regulation could impact these fundamental balancing acts between advertising revenue, creator incentives, and content variety on these platforms.

Conclusion: Ruth: So, to recap, this paper looks at how platforms manage user attention—what content gets shown versus what gets paid for—and how that affects creators and users on social media.

Warren: Yeah, it's about the revenue maximization side of things here, but also the balancing act of keeping creators from getting totally crushed by the algorithm.

Iris: From my econometric view, it’s trying to figure out which specific variables—like a creator’s inherent skill versus their current popularity—are driving that attention allocation.

Ruth: Exactly. The authors are using this framework to map out four types of creators: those who don't really get noticed, the local entertainers, the ladder climbers, and the global influencers.

Warren: And they show that under perfect information, the algorithm actually steers viewers toward content that doesn't make much money for the creator on its own.

Iris: They find that this means viewers are directed to pay too much attention to stuff where a creator’s profit margin is negative at a specific point.

Ruth: When you move into incomplete information, like when users don't tell the platform exactly what type they are, things get more complicated for monetization.

Warren: The whole structure changes because the cost for the platform to grow and make money from those network effects goes up when information is private on both sides.

Iris: And their result there is that viewer attention drops and creator effort also drops, which makes it harder to sustain the system.

Ruth: It really boils down to how we structure the incentives: are we prioritizing pure advertising revenue, or are we trying to foster a healthy ecosystem where creators keep making good stuff?

Warren: The comparison between having direct payments versus just relying on attention allocation suggests that when you let creators get paid directly, the platform might use exposure and payment together in a way that's different from just optimizing for ad clicks.

Iris: They also touch on the practical side of it, showing how quality is mapped into a promised total attention through this allocation rule.

Ruth: So, what does all this mean for regulation? It points to a trade-off: restricting how much we use personal user data in recommendations might help individual users by giving them some privacy benefits, but it could also hurt the variety of content available on the platform overall.

Warren: That's a real policy question. It forces us to think about how rules like the DMA might actually mess with this attention allocation trade-off we just talked about.

Cornell University

econ.TH

Submitted: 2026-09-14

Updated: 2026-10-28

Importance score: 80/100

The gist: We study revenue-maximizing attention allocation on an ad-funded social media platform governed by recommendation algorithms The optimal design uses attention subsidies discriminatorily.

Key concepts

Attention Subsidies: This refers to how the platform strategically directs user attention towards certain content or creators. The optimal design uses these subsidies in a discriminatory way, meaning the allocation of attention is not uniform but tailored to maximize the desired outcome, whether that's ad revenue or creator engagement.
Creator Types: The model segments creators into four distinct types: Inactive, Local Entertainers, Ladder Climbers, and Global Influencers. These types are defined by their ability to produce content and their potential for growth on the platform, which dictates how attention should be allocated to them.
Total Attention Function Xi(·): This function describes the total attention received by a creator or piece of content. The optimal allocation rule is characterized by an increasing function, meaning that as a creator's ability or potential increases, the total attention they receive also increases according to this specific mathematical relationship.
Information Rents: These are economic benefits users gain when they possess private information about their own characteristics (like their vertical type). Limiting the use of these characteristics in recommendations can potentially increase user welfare by allowing them to keep these rents, but this may come at the cost of platform diversity.

Terminology

Information Asymmetry — 20× in this paper · explained in 4 episodes
Monotonicity — 7× in this paper · explained in 2 episodes
Composition — 5× in this paper · explained in 2 episodes
Accuracy — 4× in this paper · explained in 1 episode
Core — 4× in this paper · explained in 2 episodes
Benchmark — 3× in this paper · explained in 2 episodes
Objective Function — 3× in this paper · explained in 3 episodes
Personalization — 3× in this paper · explained in 2 episodes
Sufficiency — 2× in this paper · explained in 2 episodes
Commentary — 1× in this paper · explained in 2 episodes
Convex Hull — 1× in this paper · explained in 2 episodes
EU Digital Markets Act — 1× in this paper
EU Digital Services Act — 1× in this paper
Feedback Loop — 1× in this paper · explained in 2 episodes

Summary

We study revenue-maximizing attention allocation on an ad-funded social media platform governed by recommendation algorithms The optimal design uses attention subsidies discriminatorily.

The Core Trade-off in Mechanism Design

The central economic question addressed is the trade-off between harvesting attention through advertising and cultivating it by strengthening production incentives. The platform operates as a two-sided marketplace that hosts content across multiple horizontally differentiated categories and serves viewers and creators <ref:2609.15573#pg2> Designers must balance monetization incentives against the need to provide creators with sufficient exposure to sustain the production of engaging content <ref:2609.15573#pg3>. The framework characterizes the advertisement-revenue-maximizing allocation of attention <ref:2609.15573#pg5>.

Optimal Allocation Under Complete Information

Under complete information, the optimal algorithm is characterized by an increasing total attention function Xi(·) such that xii(theta, v) = 1 if n Gij(v) >= 1 − Xi(theta) βi o, Ri(theta) = (Xi(theta) − βi)+, ei(theta) = u(Xi(theta)) <ref:2609.15573#pg4>. This characterization segments creators by ability into four types: Inactive, Local Entertainers, Ladder Climbers, and Global Influencers <ref:2609.15573#pg9>. The optimal algorithm directs viewers to pay excessive attention to content with negative pointwise surplus <ref:2609.15573#pg9>.

Incomplete Information and Network Effects

When users’ vertical types are private information, the optimal allocation is characterized by an increasing total attention function Xi(·) such that xii(theta, v) = 1 n Gij(v) >= 1 − Xi(theta) βi o, Ri(theta) = (Xi(theta) − βi)+, ei(theta) = 1/theta ∫ theta 0 u˜ dθ <ref:2609.15573#pg10>. Private information on both sides makes it more costly for the platform to expand and monetize network effects, resulting in lower viewer attention and lower creator effort.

Monetary Incentives and Comparative Statics

The model is extended to allow direct monetary payments to creators, where the optimal mechanism may use exposure and payment jointly <ref:2609.15573#pg7>. The comparative static shows that a higher z means that each unit of viewer attention is more valuable in direct advertising monetization, leading platforms to rely more on monetary transfers and less on distorted attention allocation.

Implementation and Policy Implications

The implementation involves mapping observed quality into a promised total attention via an attention allocation rule chii(q). The results suggest a regulatory trade-off: limiting the use of user characteristics in personalized recommendations may raise participating users’ welfare by leaving them information rents, but may also reduce diversity in the platform’s content ecosystem. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems. The result also has policy implications regarding how regulations like the DMA may affect the platform’s attention-allocation trade-off. The discussion points to a stark policy implication regarding category survival and reach information asymmetry. The model provides a new perspective for policy discussions on advertising regulation and recommendation systems <ref:2609.

Improvements for AI systems

Adaptive Attention Allocation based on Creator Ability: The system can dynamically adjust attention allocation to creators based on their ability level, using a function X i(theta) that dictates how much attention a creator receives, ensuring higher-ability creators receive additional cross-category attention when the marginal return to effort is large.
Network Effect Management through Discriminatory Subsidies: The AI can implement targeted subsidies by excluding low-ability creators while subsidizing high-ability ones through exposure or monetary payments, thereby endogenously determining the strength and scope of network effects.
Privacy-Aware Recommendation Filtering: By incorporating private information on both sides, the system can apply screening mechanisms that limit personalization and leave participating viewers with privacy benefits and information rents, potentially leading to a trade-off between sustaining category diversity and protecting user privacy.
Dynamic Incentive Switching via Monetary Transfers: The platform can optimally combine exposure with direct monetary transfers by choosing between them based on advertising profitability, as the model shows that platforms with stronger advertising monetization rely more on direct payments and less on distorted cross-category exposure.
Personalized Content Ordering for Viewers: The system can generate an optimal ordered feed for each viewer characterized by function K j(A), where attention A is rewarded with cumulative within-category quality weighted content Q j(v), ensuring lower-ranked content should receive less attention and have lower match quality for a given viewer.
Predictive Creator Effort Modeling: The AI can model creator effort as a function of ability and the total money-metric reward, predicting that Ladder Climbers face active trade-offs where they may increase quality in exchange for marginally further reach.