{"id":"8ecfd7b7-e0a9-4aba-8514-850214ad9c11","arxiv_id":"2504.18817","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Mastodon users in this interview study prefer chronological feeds but accept transparent, adjustable rule-based feed curation; a slider-based prototype called braids.social helped them balance home, local, and trending content.","lead":"This study interviewed 21 Mastodon users and built a prototype called braids.social that lets people mix their home, local, and trending feeds with sliders. It reports that users want chronological feeds for transparency but will accept algorithmic help when the rules are visible and adjustable.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Seamful-design claim is a causal attribution from a non-comparative interview: no baseline rules out demand characteristics, novelty, or rule-based-vs-ML confounds.","rationale":"The reader's weakest assumption is sample representativeness, and I agree that is a genuine limitation. But the more load-bearing problem is causal identification: the paper's headline design claim is that seamful design enhances acceptance, yet no comparison condition exists. Even a perfectly representative 21-user sample would not support the causal attribution in Section 6.2 because all participants experienced the same intervention. The paper's acknowledged limitations in Section 8 concern who was studied, not whether the design caused the observed attitudes. I would keep the reader's CONDITIONAL verdict: the descriptive findings about chronological preference, server choice, filtering strategies, and the prototype itself are credible and useful, but the \"enhances acceptance\" claim should be explicitly reframed as a hypothesis pending a controlled comparison. A randomized comparison of seamful vs. stripped versions is feasible and would settle the causal question; coding existing transcripts for unprompted evidence of understanding could provide a cheap internal check in the meantime. I also credit the paper for shipping a concrete prototype with pseudocode and detailed codebooks, which makes such a test easy to construct.","tokens_in":22188,"tokens_out":5394,"duration_ms":56912,"concrete_test":"Preregister a between-subjects experiment with Mastodon users (recruited from multiple servers, not the lead author's) comparing two otherwise identical versions of braids.social: (A) full seamful UI with source badges, tooltips, and immediate real-time redraw; (B) stripped control with the same sliders and feed composition, but no badges, no tooltips, and a 5-second delayed update. Randomly assign about 40 participants per condition, have both conditions perform the same three customization tasks, and measure acceptance (willingness to use, perceived transparency, trust, preference vs. chronological feed) and objective understanding (post-source identification accuracy). If condition A does not significantly outperform condition B, Sections 6.2 and 7.1's causal claim is unsupported; the qualitative data alone cannot settle it.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 6.2 states, \"We attribute this attitude difference to the transparency and user agency embedded in the design of braids.social,\" and Section 7.1 generalizes to \"seamful design enhances people's acceptance of algorithmic feed curation.\" The load-bearing assumption is that the seamful UI (source badges, sliders, real-time redraw) caused the acceptance observed in Part 2. The study is non-comparative: all ten Part 2 participants used the same fully seamful prototype, with no condition lacking badges, feedback, or explanatory cues. Part 1 and Part 2 used disjoint participant groups, and Part 1's \"algorithmic feeds\" referred to ML-based ranking while Braids is rule-based. Consequently, the reported shift from chronological preference to acceptance of algorithmic curation confounds the seamful interface with different participant pools, algorithm type, task demand, novelty, and courtesy bias in talking with the designers. Section 8 acknowledges the sample and generalizability limits but does not address this internal-validity gap. Even with a perfectly representative sample, these data cannot distinguish \"seamful design enhances acceptance\" from \"users accept rule-based curation they directly control\" or \"users reported positive impressions during a usability interview.\" The design implications in Section 7.1 therefore rest on an untested causal claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a two-part qualitative interview study with 21 Mastodon users. In Part 1, eleven participants were interviewed about how they perceive, interact with, and manage Mastodon feeds; the authors report a preference for chronological feeds and openness to algorithmic curation under conditions of transparency and user control. In Part 2, ten different participants evaluated braids.social, a web-based prototype that merges Mastodon feeds into a single semi-chronological feed with slider-based priority controls, source badges, and real-time redraw. The paper claims that seamful design enhances acceptance of algorithmic feed curation, discusses trade-offs between machine-learning-based and rule-based curation, and raises design implications for decentralized social media tools, including the choice between new apps and add-ons.","tokens_in":22413,"tokens_out":3830,"duration_ms":38552,"significance":"If the findings hold, the paper provides useful empirical grounding for an underexplored area: feed curation in decentralized social platforms. Strengths include the authenticity of participant quotes, the clear configuration table (Table 1), the transparent discussion of limitations in Section 8, and the inclusion of implementation pseudocode in Appendix B. The paper also makes a concrete design contribution with braids.social, which can serve as a reference point for future curation tools. However, the central claim about seamful design causally enhancing acceptance rests on a non-comparative study, and the Part 2 analysis lacks inter-rater reliability; these issues need to be addressed before the design implications can be considered robust.","major_comments":[{"comment":"The central claim that 'seamful design enhances people's acceptance of algorithmic feed curation' (Section 7.1) is a causal attribution based on a non-comparative usability study. The study has no baseline condition without seamful features (e.g., no badges, no real-time feedback, no explanation of the algorithm), and the ten Part 2 participants are a different group from the eleven Part 1 participants. Moreover, Part 1 participants explicitly referred to machine learning-based algorithms when discussing algorithmic feeds, while braids.social is rule-based; the authors themselves note this in Section 6.2. Consequently, the observed acceptance could be due to algorithm type, novelty, task demand, or courtesy bias, rather than to the seamful interface. Please reframe the conclusion as 'participants reported acceptance of a rule-based feed when interacting with a seamful interface' and, if the claim is to be retained, provide evidence from a within-subjects comparison that varies the seamfulness of the interface while holding the algorithm type constant.","section":"Section 6.2 and 7.1"},{"comment":"Part 1 coding used two coders with reported inter-rater reliability (Cohen's kappa ≥ 0.8), but Part 2 was analyzed by a single coder with no reliability check. This asymmetry undermines the dependability of the Part 2 themes used to support the paper's main design claims (e.g., 'badges were helpful', 'sliders were intuitive'). Please either add a second coder for a subset of Part 2 transcripts and report agreement, or explicitly add this as a limitation in Section 8.","section":"Section 3.3"},{"comment":"The combinePosts() pseudocode in Appendix B randomly selects a non-empty category for each output post, rather than deterministically ordering by the priority levels. This random interleaving is not described in the design rationale of Section 5.1, which claims that 'we retained the chronological order within each feed's fragmentation.' The random interleaving may also explain P13's confusion in Section 6.3 about trending posts not appearing immediately at the top. Please clarify the stochastic nature of the algorithm, justify it as a design choice, and discuss whether it undermines the 'semi-chronological' description and the transparency the authors claim for the tool.","section":"Section 5.3 and Appendix B"}],"minor_comments":[{"comment":"The phrase 'one export coder' appears to be a typo; it should likely be 'one expert coder'.","section":"Section 3.3"},{"comment":"Reference [52] is a self-cited workshop report from the authors' own university blog; please ensure that it is cited appropriately or replaced with a peer-reviewed source if possible.","section":"Reference list"},{"comment":"The ACM Reference Format block lists '2018' as the publication year, but the manuscript is dated 2025; this should be corrected.","section":"ACM Reference Format"},{"comment":"In the comparison list, the claim that Mastodon's non-profit nature 'grants user greater agency in curating what they want to see in their feeds' is a broad assertion not directly supported by the data; consider softening it to 'may offer' or providing a citation.","section":"Section 2.2"},{"comment":"The column 'Accounts' in Table 1 is not clearly explained in the table caption; consider renaming it to 'Prioritized Accounts' to match the terminology used in Section 5.2.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid descriptive study of an underexplored topic, and the prototype is a useful artifact. However, the main claim about seamful design causing acceptance needs a more rigorous research design or a more cautious interpretation. The single-coder analysis for Part 2 is a straightforward fix. I would encourage the authors to revise accordingly; the paper has merit for a CSCW venue once the causal inference is appropriately bounded."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: good exploratory qualitative study; the descriptive findings are the contribution, and the causal claim about seamful design should be treated as a hypothesis, not a result.\n\nThe genuinely new thing here is empirical: this is one of the first interview studies of how Mastodon users actually perceive and manage their four feeds, and the comparison of five existing clients plus the braids.social prototype extends that. Part 1 uses a solid two-coder thematic analysis with reported kappa, and the findings — chronological transparency valued, local/federated underused, server choice driven by community/moderation, filters and lists as rule-based curation — are well supported by quotes. The configuration table and pseudocode make the prototype legible. The paper is honest about sample size, self-selection, and the Google Forms/Zoom recruitment concern.\n\nThe soft spot is real and it is the one the stress-test flags. Section 6.2 says the attitude difference is \"attributed to the transparency and user agency embedded in the design,\" and Section 7.1 generalizes to \"seamful design enhances people's acceptance of algorithmic feed curation.\" The study has no comparison condition: all ten Part 2 participants used the same fully seamful prototype. There is no baseline without badges, sliders, or real-time redraw. Part 1 and Part 2 used disjoint participant groups, and Part 1's \"algorithmic feeds\" referred to ML ranking while braids is rule-based, so the acceptance shift is confounded with algorithm type, participant pool, and courtesy bias in talking to the designers. The limitations section covers generalizability but not this internal-validity gap. The right fix is to reframe 7.1 as a hypothesis with design implications, not a demonstrated effect.\n\nMinor: Part 2 coding was done by one coder with no inter-rater reliability; no code or data is shipped, so the prototype can't be inspected. These are minor compared with the causal overreach.\n\nBottom line: a useful, honest paper. The descriptive results will be cited by anyone working on Fediverse UX. It deserves a serious referee — the causal language should be softened, and the artifact should be released, but desk rejection would be a miss. Bring it to reading group if you care about decentralized social media.","headline":"A solid descriptive study of Mastodon feed curation whose causal claim about seamful design outruns its evidence.","tokens_in":22949,"tokens_out":2286,"would_cite":true,"duration_ms":23014,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Mastodon users largely prefer chronological feeds for their transparency, but they accept rule-based algorithmic curation when the interface makes the algorithm legible and keeps control in the user's hands.","keywords":["Mastodon","decentralized social media","feed curation","seamful design","algorithmic transparency","qualitative user study","chronological feed","rule-based curation"],"falsifier":"A representative study of a broader sample, for instance users of large general-purpose servers, non-English-speaking communities, and users who stick to the official app, comparing acceptance of rule-based curation with seams visible versus hidden: if a broader sample shows no increase in acceptance when the algorithm is made visible and adjustable, the paper's central claim would be refuted.","tokens_in":21971,"feed_emoji":"🐘","tokens_out":7949,"duration_ms":70920,"temperature":0.7,"pith_summary":"This paper claims that Mastodon users' attachment to chronological feeds is not a blanket rejection of algorithms: they distrust opaque, engagement-driven machine learning, but they welcome rule-based curation when the interface makes the rules legible and keeps control in the user's hands. The evidence is a two-part interview study with 21 Mastodon users, paired with braids.social, a prototype that merges the home, local, and trending feeds into one semi-chronological feed controlled by priority sliders, source badges, and real-time feedback. Participants described the unified feed as more digestible than the default multi-feed layout and credited it with improving discoverability, while accepting algorithmic mixing mainly for catching up quickly and exploring during longer absences. If this holds, it grounds concrete design guidance for decentralized feed tools: keep chronological defaults, show where each post came from, and make priority settings adjustable with immediate visible effect.","feed_headline":"Mastodon users accept algorithms—if they can see them","feed_subtitle":"Twenty-one user interviews show visible, controllable curation beats hidden ranking on Mastodon.","key_machinery":"The load-bearing mechanism is braids.social, a web-based Mastodon client built on the platform's public APIs. It collapses the home, local, and trending feeds into a single semi-chronological stream: posts are fetched from each source in proportion to slider priorities (None, Low, Medium, High), merged within per-source chronological fragments, and labeled with a badge ('Users you follow', 'Hashtag you follow', 'Trending post', 'Local post', 'Prioritized account') showing where each post came from. A 'prioritized account' field lets users pin specific accounts above the mix. The design operationalizes seamful design, a design approach that makes an algorithm's workings visible and immediately adjustable, so that user acceptance of algorithmic curation can be observed and measured.","core_discovery":"On the paper's own terms, the central discovery is that Mastodon users do not oppose algorithmic feeds as such; they oppose machine-learning feeds whose workings are hidden and whose purpose is engagement. Interviewed users prized the chronological feed because it is transparent and lets them see exactly what they missed, but they identified catching up after an absence and discovering new content as moments when algorithmic help would be welcome. When given braids.social, a unified feed in which home, local, and trending posts are mixed according to user-set priority sliders, with each post carrying a badge naming its source and the feed redrawing instantly on any change, nine of ten participants kept followed content at high priority while mixing in smaller amounts of local and trending posts. The authors conclude that seamful design, which makes the algorithm's seams visible, enhances people's acceptance of algorithmic feed curation, and that the unified feed makes information more digestible while adding discoverability. They also report that participants interpreted the slider levels high, medium, and low differently, so the legibility gains come with a need for finer granularity and stronger confirmation that the generated feed matches the settings.","pith_inferences":["The legibility lesson likely transfers to centralized platforms: source badges and adjustable priority sliders could be tested on mainstream feeds, though the profit incentive there may undermine the trust that made them work on Mastodon.","The paper tested only rule-based curation; a natural extension is to give the same badge-and-slider treatment to a machine-learning ranking and measure whether acceptance rises as much, which would separate legibility from rule-basedness as the driver.","Because recruitment was skewed toward transparency-minded users, the observed acceptance of legible algorithmic curation may be a lower bound; users with weaker transparency concerns might accept even less legible curation or might find the slider interface too demanding.","The observed multi-client fragmentation suggests a testable design direction: a plugin or add-on ecosystem layered on existing Mastodon clients, rather than another stand-alone client, may better serve users who already juggle several apps."],"forward_implications":["Feed curation tools for decentralized platforms should default to chronological order and make any algorithmic intervention visible, because transparency is what users value in the chronological feed.","Users' willingness to accept algorithmic curation tracks browsing intent: algorithmic help is wanted for quick catch-up after absence and for discovery, not for routine browsing.","Rule-based and machine-learning-based curation have distinct, irreplaceable roles, rule-based for configuring what kinds of posts to see and machine-learning for summarization and discovery, so future designs should co-design both rather than choose one.","A single unified feed with an adjustable mix of sources can reduce the under-utilization of the local and trending feeds and increase content discoverability.","Priority sliders need finer granularity and real-time confirmation, through badges and immediate redraws, to close the gap between users' mental models and the algorithm's actual behavior."],"supporting_citations":[{"why":"Documents that resistance to algorithmic feeds comes from unawareness of the algorithm, the premise that seamful design is meant to repair.","marker":"[23]"},{"why":"Documents users reasoning about invisible algorithms in news feeds, grounding the requirement that curation be transparent.","marker":"[24]"},{"why":"Supplies the validated slider-based preference interface that braids.social adapts for feed composition.","marker":"[44]"},{"why":"Motivates the prioritized-account feature by showing that users want to surface content from specific people they care about.","marker":"[30]"},{"why":"Grounds the finding that resistance to algorithmic change stems from folk theories and a perceived loss of agency.","marker":"[16]"},{"why":"Establishes that the absence of algorithmic content discovery is a known limitation for Mastodon adoption, which the machine-learning-versus-rule-based discussion addresses.","marker":"[81]"},{"why":"Supplies the thematic-analysis method used to code both interview studies.","marker":"[9]"},{"why":"Establishes Mastodon users' distinct motivations, such as safe community and social escapism, that shape feed-curation expectations.","marker":"[50]"}],"fun_headline_variants":["Mastodon users accept visible, controllable algorithms","Seamful design boosts Mastodon algorithm acceptance","Mastodon users want algorithms with visible seams","Transparent algorithms win over Mastodon users","For Mastodon users, visible beats hidden algorithms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that 21 self-selected users, recruited through the lead author's home server and interviewed over Zoom, are representative enough of Mastodon users at large to support generalizable design guidance.","fun_headline_variants_meta":{"raw":{"variants":["Mastodon users accept visible, controllable algorithms","Seamful design boosts Mastodon algorithm acceptance","Mastodon users want algorithms with visible seams","Transparent algorithms win over Mastodon users","For Mastodon users, visible beats hidden algorithms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000165,"raw_usage":{"total_tokens":1244,"prompt_tokens":936,"completion_tokens":308,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":238}},"tokens_in":552,"tokens_out":308,"duration_ms":3279,"temperature":1.0,"reasoning_tokens":238,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:08:25.632357+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A representative study of a broader sample, for instance users of large general-purpose servers, non-English-speaking communities, and users who stick to the official app, comparing acceptance of rule-based curation with seams visible versus hidden: if a broader sample shows no increase in acceptance when the algorithm is made visible and adjustable, the paper's central claim would be refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Motivates the prioritized-account feature by showing that users want to surface content from specific people they care about."}],"review_version":1}