How Often Should I Access My Online Social Networks?

Eduardo M. Hargreaves, Daniel Sadoc Menasché, Giovanni Neglia

This paper, published in 2019, provides a mathematical framework for a question we all struggle with: how often should I check my phone? By defining a new metric called "Value of Access," the authors move beyond profit-centric retention models to focus on user utility and the "attention economy."

TL;DR

The researchers propose an analytical model to determine the optimal frequency for accessing social networks. By balancing the "Value of Access" (the expected amount of new content per visit) against the attention-related costs of checking the app, they derive a closed-form expression for the ideal access rate and demonstrate how algorithmic feeds (like Facebook's) actually reduce the information value of a single visit.

What Problem Does This Paper Address?

In the "attention economy," information is abundant, but human attention is the scarce resource. Users face a dilemma: check social media too rarely, and you miss important updates; check too often, and you waste time seeing the same posts or low-quality filler.

While social media companies have studied user retention extensively to maximize ad revenue, there has been little academic focus on the user's perspective. Specifically, there was a gap in understanding the "intrinsic value" of an access. The authors aim to quantify this value and find the "sweet spot" where a user stays informed without succumbing to the diminishing returns of digital addiction or wasted effort.

Methods

The authors build a stochastic model based on three main pillars:

  1. Content Generation: They model the arrival of new posts as a Poisson process with rate $\lambda$.
  2. User Access: They model user check-ins as a process with rate $\mu$. They specifically analyze "Special Case II," where the interval between accesses is exponentially distributed.
  3. Timeline Mechanics: They assume a timeline of size $K$. If there are fewer than $K$ new posts since the last visit, the timeline is filled with "repeated" content (posts the user has already seen).

The core of the paper is the derivation of the Value of Access (VoA), defined as the expected number of novel impressions per view. Using the law of total probability and algebraic manipulation of Poisson distributions, they arrive at a compact closed-form expression for VoA.

Finally, they define a Utility Function: $U(\mu) = \text{(Access Rate)} \times \text{(VoA)} - \text{(Access Rate)} \times \text{(Cost per Access)}$ By taking the derivative of this function, they find the optimal access rate $\mu^*$.

Datasets

To validate the model, the authors used real-world data from Facebook during the 2018 Brazilian presidential elections.

  • Sources: They tracked 30 representative Brazilian media outlets and the pages of presidential candidates and their parties.
  • Collection Tool: They used the "Facebook Tracking Exposed" (FTE) browser extension.
  • Bots: They deployed virtual "bots" to simulate different user behaviors: a "high sampling" bot (checking every 10 minutes) and "regular" bots (checking every hour).
  • Trace-driven Simulation: They used the collected timestamps to simulate a chronological (FIFO) feed to compare against Facebook’s actual algorithmic News Feed.

Key Insights and Ideas

  • The Algorithmic Penalty: One of the most striking findings is that Facebook’s News Feed algorithm actually reduces the VoA compared to a simple chronological feed. Because the algorithm prioritizes "relevance" and often shows the same high-engagement posts multiple times, users see fewer new items at the top of their feed.
  • The "Scroll" Tax: The model shows that for a timeline size of $K=10$, the theoretical VoA is 10 (all new). However, on Facebook, the actual VoA was closer to 5. This means a user must scroll through roughly 20 posts to find the 10 new ones they would have seen immediately in a chronological feed.
  • Two Operating Regimes:
    • Content-Limited: If you check the app rarely, your VoA is limited by the timeline size ($K$). You are likely missing content that was pushed out of the timeline before you arrived.
  • Access-Limited: If you check very frequently, your VoA is limited by the rate of content production ($\lambda$). You are likely seeing the same posts over and over.
  • Optimal Frequency is Non-Linear: The optimal access rate $\mu^*$ doesn't just increase linearly with content production; it reaches a point of diminishing returns where the "cost" of the access (attention/time) outweighs the tiny bit of new information gained.

Key Results

  • Mathematical Proof: The authors successfully derived a closed-form expression for VoA: $\bar{V} = \frac{\lambda}{\mu} (1 - (\frac{\lambda}{\lambda + \mu})^K)$.
  • Model Accuracy: Trace-driven simulations showed that the Poisson process is a "good approximation" for real-world content generation, despite the inherent "burstiness" of news cycles.
  • Position Bias: The data confirmed a strong "position bias" in the News Feed, where repeated content is intentionally placed at the top to ensure engagement, even if it lowers the information density of the access.

Limitations and Future Work

  • Static Cost: The model assumes a fixed cost $c$ per access. In reality, the cost might increase if a user decides to scroll deeper (increasing $K$) to find more new content.
  • Bursty Content: While Poisson works well for averages, it doesn't account for "bursts" of news (e.g., during a debate). The authors suggest moving toward an $M/G/K$ queueing model with bulk services in future iterations.
  • Subjective Value: The model treats all "novel" posts as having equal value. Future work could incorporate the quality or relevance of the content into the VoA metric.
  • Platform Comparison: The study was limited to Facebook; applying this to high-velocity platforms like X (Twitter) or low-velocity platforms like LinkedIn could yield different optimal behaviors.

Summary updated

Abstract

Users of online social networks are faced with a conundrum of trying to be always informed without having enough time or attention budget to do so. The retention of users on online social networks has important implications, encompassing economic, psychological and infrastructure aspects. In this paper, we pose the following question: what is the optimal rate at which users should access a social network? To answer this question, we propose an analytical model to determine the value of an access (VoA) to the social network. In the simple setting considered in this paper, VoA is defined as the chance of a user accessing the network and obtaining new content. Clearly, VoA depends on the rate at which sources generate content and on the filtering imposed by the social network. Then, we pose an optimization problem wherein the utility of users grows with respect to VoA but is penalized by costs incurred to access the network. Using the proposed framework, we provide insights on the optimal access rate. Our results are parameterized using Facebook data, indicating the predictive power of the approach.
BibTeX
@inproceedings{hargreaves2019how,
  author = {Eduardo M. Hargreaves and Daniel Sadoc Menasché and Giovanni Neglia},
  title = {{How Often Should I Access My Online Social Networks?}},
  booktitle = {MASCOTS},
  year = {2019},
  url = {https://openreview.net/forum?id=ocE6RnJkkZL},
}

Loading related content...

Figure