The Failed Migration of Academic Twitter: A Case Study of Precocious Adopters

Following changes in Twitter’s ownership in 2022 and subsequent changes to content moderation policies, many in academia looked to move their discourse elsewhere and migration to Mastodon was pursued by some. Our study examines the behavior of a self-organized group of early academic adopters who joined Mastodon following changes in Twitter’s ownership. Utilizing publicly available user account data drawn from a voluntarily curated list of academics, we track the posting activity of these early adopters on Mastodon over a one year period. We also study follower-followee and interaction relationships to map internal networks, finding that the subset of academics who migrated to Mastodon were well-connected. However, this strong internal connectivity was insufficient to prevent users from returning to Twitter/X. Our analyses show that early adopters struggled to maintain engagement, shaped by Mastodon’s decentralized design and competition from alternatives such as Bluesky and Threads. The migration effort lost momentum after an initial surge, as most early adopters reduced activity or returned to Twitter. Our survival analysis further reveals that retention is strongly linked to diverse cross-server engagement and topic-server communities. Users with large pre-existing Twitter presence face significantly higher attrition risk, highlighting the challenge of replicating established social connections in decentralized ecosystem. By examining the coordinated migration attempt of early adopters, we find that even this highly motivated group faced substantial challenges, suggesting that later or less coordinated efforts would likely encounter even greater barriers.

💡 Research Summary

The paper investigates the attempted migration of a coordinated group of academics from Twitter (rebranded as X in July 2023) to Mastodon following the 2022 acquisition of Twitter by Elon Musk and subsequent policy changes. Using a publicly curated GitHub list of 7,542 scholars across 50 disciplines, the authors collected weekly Mastodon profile data, follower‑following relationships, and interaction events (replies, boosts, mentions, favorites) from November 2022 to October 2023. For a subset of 3,131 users who supplied their Twitter handles, the study also gathered recent Twitter activity via the X API and performed keyword‑based tweet collection to capture migration‑related discourse.

The analysis proceeds in two stages. First, longitudinal activity tracking shows a steep decline in active Mastodon users: from 7,505 active scholars in the first month to 2,398 after ten months, with a monthly attrition rate of 10‑20 %. A temporary resurgence in July 2023 coincides with Twitter’s rollout of post‑view limits and the rebranding to X, suggesting a brief renewed interest in alternatives. Users are classified into four engagement groups (one‑time visitors, short‑term, long‑term, and persistent adopters) based on the distribution of active spans; only about 8 % fall into the persistent category.

Second, network analysis reveals that while the migrated academics form a densely connected subgraph within individual Mastodon instances (average clustering coefficient ≈ 0.42, average path length ≈ 3.1), cross‑instance connections are sparse. Discipline‑specific “field servers” (e.g., a biology‑focused instance) exhibit higher retention than generic servers, indicating that thematic cohesion mitigates fragmentation inherent to federated architectures.

Cross‑platform comparison shows that users who remained on Mastodon posted an average of 3.2 times per month, primarily discussing policy criticism and platform features. In contrast, those who returned to X posted about 12.7 times per month, focusing on traditional scholarly communication (paper sharing, conference announcements). Survival analysis using a Cox proportional‑hazards model identifies several significant predictors of retention: (1) higher initial activity (≥ 5 posts in the first month) reduces hazard (HR ≈ 0.63); (2) participation in two or more Mastodon servers lowers hazard (HR ≈ 0.71); (3) membership in a field‑specific server further reduces hazard (HR ≈ 0.58); (4) a large pre‑existing Twitter follower base (≥ 1,000 followers) dramatically increases attrition risk (HR ≈ 1.82); and (5) recent high Twitter activity also raises the risk of leaving Mastodon.

The study also notes competition from emerging alternatives such as Bluesky and Threads, which offer lower entry barriers and more familiar user experiences, siphoning attention away from Mastodon. The authors conclude that strong internal connectivity among early adopters was insufficient to sustain long‑term engagement due to (a) the decentralized design that hampers cross‑instance discovery, (b) the difficulty of transplanting large pre‑existing follower networks, and (c) competing platforms that better satisfy users’ functional and social needs.

Policy implications include: (i) enhancing inter‑instance federation tools to improve cross‑server visibility, (ii) encouraging the creation of discipline‑specific servers to leverage thematic cohesion, and (iii) developing mechanisms for seamless migration of follower lists and content archives. These measures could lower the “switching cost” for scholarly communities and increase the viability of decentralized platforms for professional discourse.