Pith. sign in

REVIEW 3 major objections 5 minor 34 references

Deterrence Effects of Social Media Interventions on Health Misinformation Dissemination by Bots and Humans

T0 review · 3 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Platform interventions durably reduced bot accounts and bot activity in cancer misinformation over a decade, while human sharers showed no significant sustained change.

desk verdict The bot deterrence finding is likely an artifact of sample attrition; the human null is the real contribution, but the paper overstates the bot side. read the letter →

arxiv 2607.18248 v1 pith:2ZFGPAHL submitted 2026-05-08 cs.HC

Amir Karami , Nima Kordzadeh , Serena Harn This is my paper · ORCID
classification cs.HC
keywords healthmisinformationsocialmediainterventionsbotsdeterrencetheoryinterruptedtimeseriesARIMAcancerTwitter
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that Twitter's three anti-misinformation interventions—removal, reduction, and informing—produced lasting deterrence for automated bot accounts but not for human accounts. Analyzing tweets endorsing unproven cancer claims from 2011 to 2021, the authors find sustained declines in bot counts after all three interventions and in bot activity after removal and reduction, with estimated drops of 72–98%. For human users, no intervention produced a statistically significant sustained change in either the number of sharers or their activity. The point of the distinction is practical: aggregate studies can make interventions look successful when the effect is concentrated in bots, masking unchanged human behavior that may need different strategies.

What carries the argument

An interrupted time-series design using seasonal ARIMA models, in which a step-change indicator switches from 0 to 1 at the month each intervention was announced (February 2017 for removal, June 2018 for reduction, September 2020 for informing). The coefficient on that indicator is the estimated sustained level shift in logged monthly bot/human counts and activity. Around this, the analysis relies on identifying supportive misinformation tweets with a machine-learning classifier and separating accounts into bots and humans with a bot-scoring tool; the estimated effect percentages are derived by back-transforming the intervention coefficients.

What would settle it

Recompute bot counts and activities while imputing bot scores for the 33.1% of unscored accounts (suspended, non-existent, protected, tweet-removed), or track account status over time; if the post-intervention bot decline disappears or shrinks to non-significance once those accounts are accounted for, the sustained deterrence claim for bots is an artifact of missing-data mechanics.

Watch

Extended reading notes

Core claim

The central claim is that social media interventions act as deterrent mechanisms with different long-term trajectories for bots and humans. Using monthly counts of accounts and posts supporting cancer misinformation, interrupted time-series models with a step-change indicator for each intervention show that bot user numbers fell significantly after removal, reduction, and informing interventions, and bot activity fell after removal and reduction; the informing intervention did not significantly reduce bot activity. The same models found no significant sustained effect on the number of human sharers or the volume of human misinformation activity for any intervention. A post hoc aggregate anal

Load-bearing premise

The analysis assumes that bot and human counts are measured consistently before and after each intervention, even though a third of accounts could not be scored because they were suspended, deleted, protected, or had their tweets removed — and the removal intervention itself suspends accounts.

Editorial extensions

If this is right

  • If the paper is right, platform enforcement can durably shrink the automated side of health misinformation, so bot-oriented harms like scale amplification and distorted popularity can be reduced for years.
  • The null results for humans imply that removal, reduction, and informing as deployed on Twitter do not, by themselves, durably change human sharing of health misinformation.
  • Evaluations that mix bots and humans can misattribute bot reductions to overall success; disaggregated measurement is needed before concluding an intervention worked.
  • Since informing alone did not reduce bot activity even though it reduced bot counts, persuasive or soft interventions may be weaker than coercive or situational ones for curbing bot output.
  • The findings suggest platform resources should be split: automated-account enforcement for bots, and different, perhaps cognitive or prebunking, strategies for humans.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: A large share of accounts (about 33%) had no bot score because they were suspended, deleted, protected, or scrubbed; if those missing accounts skew toward bots, the post-removal decline in measured bot counts is partly mechanical — the intervention removes accounts from observation — rather than evidence that remaining or future bot operators were deterred.
  • Editorial inference: The paper's logic for bots implies a testable prediction: bot creators' behavior should shift toward evasion, so one would expect detection-resistant bot designs or migration to other platforms to appear after each intervention; the paper does not test this.
  • Editorial inference: The same step-change design could be applied to other persistent health misinformation topics or to newer LLM-based bots, which the authors flag as future work; re-running the analysis on post-2022 data would tell whether the bot-deterrence pattern survives the arrival of generative AI.
  • Editorial inference: The informing intervention's significant effect on bot counts but not bot activity hints that labels mainly suppress account creation or existence, while active bots that remain keep posting at similar rates; this asymmetry could be checked by stratifying bot activity by account age.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper uses Twitter data from 2011–2021 on cancer misinformation, applies ML classification to identify supportive tweets, scores accounts with Botometer, and estimates seasonal ARIMA interrupted time-series models to test whether removal (Feb 2017), reduction (Jun 2018), and informing (Sep 2020) interventions have sustained deterrent effects on the number of bots/humans and their misinformation activities. The results show significant post-intervention declines for bots (all three interventions for bot count; removal and reduction for bot activity), whereas no significant effects are found for humans. The paper concludes that platform interventions durably suppress automated misinformation spread but leave human sharing unchanged.

Significance. If the bot-related findings were valid, the paper would make a meaningful contribution: it is one of the few longitudinal evaluations that disaggregate intervention effects by user type, and it uses a decade of real-world data. The human-null result is interesting and is less vulnerable to the missing-data problem. However, the headline bot claim is compromised by a load-bearing selection issue: 33.1% of accounts received no Botometer score, and the removal intervention operates by suspending accounts, so the post-intervention bot decline may be partly mechanical. The paper's strengths include its large dataset, the explicit post hoc aggregated comparison, and the longitudinal design, but the central bot conclusion is not established as written.

major comments (3)
  1. [§4.4, Tables 4–5] The bot series are computed only for accounts with Botometer scores; 33.1% of the 96,836 supportive accounts are unscorable, including 14.2% suspended and 13.3% non-existent. Because the removal intervention operates by suspending accounts, the post-Feb-2017 decline in bot counts/activities may reflect mechanical attrition from the measurement frame rather than deterrence. The paper never reports missingness by month or user type, nor does it compare the scored bot series with the total supportive-account series. Please add missingness diagnostics, attempt bounding/robustness checks, and reframe the bot conclusions accordingly. As written, H1a–H1c and H2a–H2b are not established.
  2. [§4.5, Appendix A, Tables 4–7] The ARIMA specifications are not reported: the orders ((p,d,q)×(P,D,Q)), the exact intervention coding (separate vs. joint models), and the estimation details are missing. Because the three interventions overlap in time (removal Feb 2017, reduction Jun 2018, informing Sep 2020), separate step models confound later interventions with earlier accumulated effects. Report full model formulas, estimation output, and sensitivity to alternative intervention start dates.
  3. [Table 4, Table 5] The estimated effects are extremely large: removal implies an 84% decrease in bot count and a 98% decrease in bot activities. These magnitudes are hard to reconcile with the reported 14.2% suspended-account share even if all suspended accounts were bots. The gap suggests either that the measured decline is dominated by attrition of newly created bots or that the model is misspecified; the paper should reconcile these numbers or provide a mechanism.
minor comments (5)
  1. [§5, Table 5 paragraph] The text says 'supporting H2a and H2b, but not H3c'; H3c is a human hypothesis. This should be H2c.
  2. [§4.3, Table 3] The total number of tweets is inconsistent: the text reports 268,280 collected, 2,000 used for training, and 266,280 unlabeled, but Table 3 lists 264,280 unlabeled and a total of 266,280. Verify and correct the arithmetic.
  3. [Appendix B, Table B.2] Table B.2 appears to be a copy of Table 7 (same estimates and model assessment), yet the text describes it as modeling the number of all users. Clarify the intended analysis and correct the table.
  4. [§6.3 Limitations] The limitations section does not mention the 33.1% of accounts without Botometer scores or the direct relationship between the removal intervention and the measurement of bot presence. This omission should be addressed explicitly.
  5. [Figure 2] The caption refers to 'the introduction of Twitter’s intervention measures' but the figure appears to mark only one date; specify which intervention the vertical line denotes.

Circularity Check

1 steps flagged · score 5.0 of 10

Bot deterrence estimates are partly built from sample attrition: suspended accounts are excluded from the bot outcome by construction.

  1. self definitional [Section 4.4 (Bot score measurement); Section 5, Table 4]
    "Among the 96,836 accounts endorsing cancer misinformation, 66.9% received scores between 1 and 5, while the rest were suspended (14.2%), non-existent (13.3%), protected (4.5%), or had their tweets removed (1.1%). ... The deterrence parameters for the removal, reduction, and informing strategies were -0.804, -0.892, and -0.554, respectively, all statistically significant."

    The outcome 'bot users' is computed only from accounts that received Botometer scores; accounts suspended by the removal intervention fall into the excluded 'suspended' category. Since the removal intervention began in February 2017 and works by suspending accounts, the post-intervention decline in bot users/activities is partly a mechanical consequence of the measurement frame dropping the very accounts the intervention removed, rather than evidence of deterrence among remaining bots. The paper does not report how missingness varies by month or user type, so the bot-versus-human asymmetry is not separated from sample attrition.

full rationale

This is an empirical interrupted-time-series study rather than a derivation, so most derivation-circularity modes do not apply. The ARIMA modeling with step intervention dummies does not reintroduce the hypotheses as inputs. The one load-bearing construction issue is in the bot measurement: Botometer scores are available for only 66.9% of the 96,836 supportive accounts; the remaining 33.1% are suspended, non-existent, protected, or tweet-removed. Because the removal intervention operates by suspending accounts, those accounts drop out of the measurable bot series after February 2017. The ARIMA removal coefficients for bot users and bot activities are therefore partly mechanical attrition from the measurement frame, not evidence about deterrence among the bots that remain observable. Since the paper's headline asymmetry is 'bots decline, humans do not,' and missingness is not analyzed by period or user type, the bot-specific conclusion is not fully identified. The human null result and the reduction/informing results have more independent content, and the self-citations (e.g., Karami 2025 typology) are background rather than load-bearing. Hence a moderate partial-circularity score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central estimates rest on measurement choices: which accounts receive bot scores, which dates mark interventions, and which ARIMA orders are selected. The most consequential is the exclusion of accounts without Botometer scores, which can mechanically produce post-removal declines in bot counts.

free parameters (3)
  • ARIMA model orders (p,d,q)×(P,D,Q) = not reported
    Chosen per series to achieve stationarity and fit; not disclosed in text or Appendix A, yet the intervention coefficients and p-values depend on them.
  • Intervention start months = Feb 2017; Jun 2018; Sep 2020
    Derived from Twitter blog announcements; treating the announcement date as the abrupt onset of a sustained step change is a modeling choice.
  • Botometer score threshold window = 2.5 to 4 (averaged in 0.1 steps)
    To avoid committing to one threshold, counts are averaged over thresholds; this is a defensible robustness choice but not a single value, and the paper does not report threshold-specific results.
assumptions (5)
  • domain assumption Twitter blog announcement dates mark the actual onset of each intervention
    Section 4.5 uses these dates as T0 for the step indicator; actual enforcement rollout was gradual and unobserved.
  • domain assumption Botometer scores reliably distinguish bots from humans
    Section 4.4; Botometer error is known to vary, and 33.1% of accounts have no score and are excluded from the analysis.
  • domain assumption Google Fact Check API entries are a valid enumeration of cancer misinformation claims
    Section 4.1; all queries are built from these 81 claims, so any incompleteness in the fact-check database defines the scope of the outcome.
  • domain assumption Classifier labels are accurate enough for temporal comparison
    Section 4.3; the 79.75% overall accuracy is not validated across time periods or against bot/human subsets; misclassification could vary with intervention periods.
  • domain assumption A step-change ARIMA captures the sustained deterrence effect
    Section 4.5 and Appendix A; the model assumes a permanent level shift at T0 and does not test slope changes or delayed/gradual effects.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deterrence Effects of Social Media Interventions on Health Misinformation Dissemination by Bots and Humans." pith.science (2026). https://pith.science/paper/2ZFGPAHL

@misc{pith2026260718248,
  author       = {Pith},
  title        = {Pith review of: Deterrence Effects of Social Media Interventions on Health Misinformation Dissemination by Bots and Humans},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2ZFGPAHL}},
  note         = {Machine review of arXiv:2607.18248}
}
read the original abstract

In the realm of social media, information dissemination is pivotal, yet it is tainted by the proliferation of misinformation propagated by both bots and humans, bearing consequential impacts on individuals and society. To address this issue, social media platforms have implemented removal, reduction, and informing interventions, acting as deterrent mechanisms to dissuade users from engaging in the spread of misinformation. Nonetheless, the sustained effectiveness of these interventions on bots and humans remains unclear. Drawing on deterrence theory, this study examines the efficacy of social media interventions on bots and humans sharing health misinformation. Our results show that most interventions can have sustained effects on bots and their activities for years after intervention implementation. However, the interventions may not have significant deterrence effects on humans and their activities. Our findings offer important theoretical and practical implications by highlighting the importance of studying both bots and humans and developing creative strategies to tackle health misinformation dissemination.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 5 canonical work pages

  1. [1]

    Metrics such as accuracy, F-measure, precision, and recall were employed to compare the algorithms and identify the most effective features

    involved a five-fold cross-validation approach, wherein tweets were divided into five subsets, with one set aside for testing and the remaining four used for training. Metrics such as accuracy, F-measure, precision, and recall were employed to compare the algorithms and identify the most effective features. As a result, each tweet was labeled as supportiv...

    2012

  2. [3]

    Basic Statistics of Categorizers. Data Total Relevant Irrelevant Supportive Other Training Data 2000 1249 299 452 Unlabeled Data 264,280 180,074 26,334 57,872 Total 266,280 181,323 26,633 58,324 Tweets Feature Extraction Unigrams Bigrams Trigrams Training Data Classification Dimensionality Reduction Test Data Unlabeled Tweets 4.4 Bot score measurement To ...

    2000

  3. [4]

    Reopening America

    https://doi.org/10.51685/jqd.2024.icwsm.7 Yang, K.-C., Varol, O., Davis, C. A., Ferrara, E., Flammini, A., & Menczer, F. (2019). Arming the public with artificial intelligence to counter social bots. Human Behavior and Emerging Technologies, 1(1), 48–61. Young, V. A. (2020). Nearly Half of the Twitter Accounts Discussing “Reopening America” May Be Bots. h...

    arXiv 2024

  4. [5]

    To address this, we computed means for bots and humans, including their tweets and retweets, across thresholds from 2.5 to 4 in increments of 0.1

    but increases the risk of false positives (Gallwitz & Kreil, 2022), while higher thresholds may increase the risk of false negatives. To address this, we computed means for bots and humans, including their tweets and retweets, across thresholds from 2.5 to 4 in increments of 0.1. On average, humans constituted 51.5% of users and contributed 40.9% of activ...

    2022

  5. [7]

    The results revealed that social media interventions had significant and lasting deterrence effects on bots, while similar effects were not observed in the case of humans

    Conclusion This study represents the initial endeavor to examine the immediate and sustained deterrence effects of social media interventions on both bots and humans engaged in the dissemination of health misinformation. The results revealed that social media interventions had significant and lasting deterrence effects on bots, while similar effects were ...

    arXiv 2023

  6. [8]

    Supported H1b The reduction intervention strategy has a sustained deterrence effect on the number of bot users spreading health misinformation

    Summary of Hypothesis Testing Results Hypothesis Hypothesized Effect Result H1a The removal intervention strategy has a sustained deterrence effect on the number of bot users spreading health misinformation. Supported H1b The reduction intervention strategy has a sustained deterrence effect on the number of bot users spreading health misinformation. Suppo...

    2019

  7. [10]

    allows misinformation to continue shaping attitudes, which limits the long-term impact of platform responses. Third, cognitive tendencies such as the truth-default theory (i.e., believing that others tell the truth more often) (Levine, 2014), naïve realism (i.e., believing that one’s current views are the only correct views) (Shu et al., 2020), and confir...

    2014

  8. [11]

    further undermine individuals’ ability to critically evaluate and resist misinformation. Fourth, interventions (e.g., suspending malicious actors) may redirect rather than eliminate human participation, as suspended or blocked users often migrate and repost content on other platforms (Horta Ribeiro, Hosseinmardi, et al., 2023), as seen when blocked commen...

    2023

Show all 34 references
  1. [12]

    https://doi.org/10.1186/s40493-014-0012-y Bluesky. (2024). Community Guidelines. Bluesky. https://bsky.social/about/support/community-guidelines Boehm, L. E. (1994). The validity effect: A search for mediating variables. Personality and Social Psychology Bulletin, 20(3), 285–2...

    2024 doi

  2. [13]

    as an AI language model

    may offer more effective pathways for curbing health misinformation among human users. Our findings underscore the importance of distinguishing between bots and humans. Bots and humans behave differently, and without separating these groups in analysis, the results can be misl...

    2024

  3. [25]

    F., & Shin, E

    https://www.demandsage.com/twitter-statistics/#:~:text=Let%20us%20take%20a%20closer,528.3%20million%20monthly%20active%20users Shin, D., Kee, K. F., & Shin, E. Y. (2023). The Nudging Effect of Accuracy Alerts for Combating the Diffusion of Misinformation: Algorithmic News Sour...

    2023

  4. [28]

    https://scholarspace.manoa.hawaii.edu/items/30ebd144-6790-4abc-bb0e-c1896009b0c9 Vasconcelos Silva, C., Jayasinghe, D., & Janda, M. (2020). What can Twitter tell us about skin cancer Communication and prevention on social media? Dermatology, 236(2), 81–89. Vincent, E. M., Thér...

    2020

  5. [33]

    Tran, T., Valecha, R., & Rao, H. R. (2023). Machine and human roles for mitigation of misinformation harms during crises: An activity theory conceptualization and validation. International Journal of Information Management, 70, 102627. Truong, B. T., Kim, S., Nogara, G., Verdo...

    2023

  6. [53]

    https://doi.org/10.1186/s41235-024-00582-6 Shewale, R. (2023). Twitter Statistics In

    2023 doi

  7. [63]

    inoculation

    Pham, B. T., Bui, D. T., & Prakash, I. (2017). Landslide susceptibility assessment using bagging ensemble based alternating decision trees, logistic regression and J48 decision trees methods: A comparative study. Geotechnical and Geological Engineering, 35(6), 2597–2611. Porte...

    2017

  8. [93]

    K., Frischlich, L., & Lermer, E

    https://doi.org/10.1186/s12939-025-02451-0 Koch, T. K., Frischlich, L., & Lermer, E. (2023). Effects of fact‐checking warning labels and social endorsement cues on climate change fake news credibility and engagement on social media. Journal of Applied Social Psychology, 53(6),...

    2023

  9. [104]

    L., Varol, O., Yang, K.-C., Flammini, A., & Menczer, F

    Shao, C., Ciampaglia, G. L., Varol, O., Yang, K.-C., Flammini, A., & Menczer, F. (2018). The spread of low-credibility content by social bots. Nature Communications, 9(1), 1–9. Sharma, P. R., Spearing, E. R., Wade, K. A., & Jobson, L. (2024). Distress reactions and susceptibil...

    2018

  10. [130]

    Karami, A., Zain, A., & Jamal, A. (2025). Unveiling the information mirage: A systematic literature review of health misinformation on social media. Journal of Public Health. https://doi.org/10.1007/s10389-025-02639-2 Katzowitz, J. (2018, August 6). This grandmother tweets so ...

    2025 doi

  11. [221]

    Google. (n.d.). Misinformation policies (Youtube). Retrieved June 25, 2025, from https://support.google.com/youtube/answer/10834785?hl=en Hagen, L., Neely, S., Keller, T. E., Scharf, R., & Vasquez, F. E. (2022). Rise of the Machines? Examining the Influence of Social Bots on a...

    2025

  12. [365]

    A., Forstner, S., Glance, J., Green, G., Kawata, A., Kovvuri, A., Martin, J., & Morgan, E

    Clayton, K., Blair, S., Busam, J. A., Forstner, S., Glance, J., Green, G., Kawata, A., Kovvuri, A., Martin, J., & Morgan, E. (2020). Real solutions for fake news? Measuring the effectiveness of general warnings and fact-check tags in reducing belief in false stories on social ...

    2020

  13. [545]

    Wilner, T., & Holton, A. (2020). Breast cancer prevention and treatment: Misinformation on Pinterest,

    2020

  14. [875]

    https://doi.org/10.1016/S1473-3099(20)30565-X Théro, H., & Vincent, E. M. (2022). Investigating Facebook’s interventions against accounts that repeatedly share misinformation. Information Processing & Management, 59(2), 102804. Thorson, E. (2016). Belief echoes: The persistent...

    2022 doi

  15. [1069]

    Papakyriakopoulos, O., Serrano, J. C. M., & Hegelich, S. (2020). The spread of COVID-19 conspiracy theories on social media and the effect of content moderation. Harvard Kennedy School Misinformation Review, 1(3). Pennycook, G., & Rand, D. G. (2019). Fighting misinformation on...

    2020

  16. [1978]

    The test helps determine whether the observed autocorrelation in the data differs significantly from what would be expected under the assumption of no autocorrelation

    to assess ARIMA models. The test helps determine whether the observed autocorrelation in the data differs significantly from what would be expected under the assumption of no autocorrelation. The test checks if errors in the data randomly bounce around, like coin flips, or if ...

    2005

  17. [1984]

    Non-stationarity in time series data can stem from two primary sources

    indicated that the analyzed time series was non-stationary. Non-stationarity in time series data can stem from two primary sources. The first is varying variance over time, known as heteroscedasticity, which can often be mitigated by applying a logarithmic transformation. The ...

    2018

  18. [1993]

    by contextualizing its three common forms, namely hard deterrence (coercive), situational deterrence (restrictive), and soft deterrence (persuasive), within the domain of social media misinformation and interventions. Although these forms have traditionally been applied in are...

    2016

  19. [1998]

    ITSA allows for multiple measurements before and after an intervention, enabling the exploration of longitudinal trends in intervention effects

    to test our hypotheses. ITSA allows for multiple measurements before and after an intervention, enabling the exploration of longitudinal trends in intervention effects. We used the autoregressive integrated moving average (ARIMA) model (Box & Jenkins, 1976), which effectively ...

    1976

  20. [2016]

    A threshold of 2.5 is common (Bessi & Ferrara,

    to 4 (Broniatowski et al., 2018). A threshold of 2.5 is common (Bessi & Ferrara,

    2018

  21. [2018]

    I agree with you, bot!

    American Journal of Public Health, 110(S3), S300–S304. https://doi.org/10.2105/AJPH.2020.305812 Wischnewski, M., Ngo, T., Bernemann, R., Jansen, M., & Krämer, N. (2024). “I agree with you, bot!” How users (dis)engage with social bots on Twitter. New Media & Society, 26(3), 150...

    2020

  22. [2019]

    to assess the likelihood of an account being a bot. The features used by Botometer are from six groups: user metadata (e.g., number of followers), friends (e.g., distribution of friends), content (e.g., number of words in a tweet), sentiment (e.g., number of negative emotions ...

    2017

  23. [2020]

    showed no significant upward or downward trend from July 2011 through December 2021, indicating stable search behavior for “Cancer” over a decade. Figure

    2011

  24. [2023]

    ) takes the value 0 when the intervention is absent (before T₀) and switches to 1 when the intervention is present (after T₀) (Schaffer et al., 2021): 𝑆

    and celebrity endorsements (e.g., skin cancer (Vasconcelos Silva et al., 2020)), which can also contribute to fluctuations in social media conversations about cancer. To account for these seasonal effects, we applied a seasonal ARIMA model. This study focused on the sustained ...

    2020

  25. [2159]

    https://doi.org/10.3390/ijerph18042159 Karami, A., Qiao, Z., Zhang, X., Kharrazi, H., Bozorgi, P., & Bozorgi, A. (2024). Health Use Cases of AI Chatbots: Identification and Analysis of ChatGPT Prompts in Social Media Discourses. Big Data and Cognitive Computing, 8(10),

    2024 doi

  26. [3284]

    Zhong, W., Broniatowski, D., Dredze, M., & Abroms, L. (2023). Evaluating Twitter’s COVID-19 Vaccine Misinformation Removal Policy. https://osf.io/preprints/socarxiv/cxg6y/ Appendices Appendix A Statistical Analysis We used a seasonal ARIMA model expressed as (𝑝,𝑑,𝑞)× (𝑃,𝐷,𝑄)*....

    2023

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.