CoChat
Sign Up
Cited by 1,700Mental Health via WritingAuthorship Attribution and ProfilingPersonality Traits and Psychology

Personality, Gender, and Age in the Language of Social Media: The Open-Vocabulary Approach

Open access
Schwartz, H. Andrew, Eichstaedt, Johannes C., Kern, Margaret L., Dziurzynski, Lukasz, Ramones, Stephanie M., Agrawal, Megha, Shah, Achal, Kosinski, Michal +3 more·2013·PLoS ONE
DOI 10.1371/journal.pone.0073791PMC PMC3783449
Open this paper in CoChat — free account

Abstract

We analyzed 700 million words, phrases, and topic instances collected from the Facebook messages of 75,000 volunteers, who also took standard personality tests, and found striking variations in language with personality, gender, and age. In our open-vocabulary technique, the data itself drives a comprehensive exploration of language that distinguishes people, finding connections that are not captured with traditional closed-vocabulary word-category analyses. Our analyses shed new light on psychosocial processes yielding results that are face valid (e.g., subjects living in high elevations talk about the mountains), tie in with other research (e.g., neurotic people disproportionately use the phrase 'sick of' and the word 'depressed'), suggest new hypotheses (e.g., an active life implies emotional stability), and give detailed insights (males use the possessive 'my' when mentioning their 'wife' or 'girlfriend' more often than females use 'my' with 'husband' or 'boyfriend'). To date, this represents the largest study, by an order of magnitude, of language and personality.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper introduces differential language analysis (DLA), an open-vocabulary method that correlates words, phrases, and LDA topics from Facebook status updates with gender, age, and Big Five personality. Across roughly 75,000 volunteers and 700 million language-feature instances, open-vocabulary features yielded psychosocial insights and outperformed LIWC in predictive models, reaching 91.9% gender classification accuracy.

Key findings

  • Open-vocabulary features outperformed LIWC in out-of-sample prediction for gender, age, and personality. Best gender accuracy was 91.9% with WordPhrases + Topics versus 78.4% for LIWC; best age \(R\) was .84 versus .65 for LIWC; best personality \(R\) was .42 for openness versus .29 for LIWC.
  • DLA produced face-valid and novel language correlations: high-elevation subjects talked about “the mountains” while low-elevation subjects talked about “beach”; neurotic people disproportionately used “sick of” and “depressed”; emotional stability was linked to social activities such as sports, vacation, beach, church, and team; introverts showed interest in Japanese media such as anime, manga, and Japanese emoticons.
  • Gender and age patterns included females using more emotion words, first-person singulars, and social/psychological process words, while males used more swear words and object references; younger groups used more slang, emoticons, and Internet speak, with topics progressing from school to college to work to family, and “we” increasing after age 22 while “I” decreased.
  • Males preceded references to their opposite-sex partner with “my” significantly more often than females did for “husband” or “boyfriend,” e.g., “my wife” and “my girlfriend” were strongly male-correlated.

Methods

  • Data came from Facebook status updates volunteered through the My Personality application. Analyses were restricted to users indicating English as a primary language, writing at least 1,000 words, being under 65, and providing gender and age; the final sample was about 75,000 volunteers, 15.4 million status updates, 309 million words, and 700 million feature instances. Personality was measured with the International Personality Item Pool proxy for the NEO-PI-R.
  • DLA extracted words and phrases (n-grams of size 1–3 passing a collocation filter) and LDA topics. Feature counts were normalized by total word use and Anscombe-transformed; each feature was correlated with outcomes using ordinary least squares regression, adjusting for gender and age, with significance at a Bonferroni-corrected two-tailed \(p < 0.001\).
  • Predictive evaluation used 75% training and 25% test data, with a linear SVM for gender and ridge regression for age and personality; features were reduced via principal component analysis and regularization was set by validation. Models compared LIWC, topics, words/phrases, and combinations.

Limitations

  • The paper states that the work is subject to sampling and social desirability biases, that language connections with psychosocial variables are context-dependent and may differ under other contexts, and that older individuals are rare in social media, limiting fine-grained lifespan tracking.
  • The paper also states that open-vocabulary analysis is not a complete replacement for closed-vocabulary approaches: with a specific theory or small sample, a priori word lists can be ideal, concepts can be drowned out by more predictive concepts, and static categories are easier to compare across studies.
  • Evident from the text: the sample is self-selected volunteers, restricted to English, under 65, and users with at least 1,000 words, with a young mean age of 23.43, so findings may not generalize to non-Facebook, non-English, older, or less verbose populations.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pone.0073791

Cites

97 of 97 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 65 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Neurotic people disproportionately use the phrase 'sick of' and the word 'depressed'.

    “neurotic people disproportionately use the phrase ‘sick of’ and the word ‘depressed’”
    Quote in the paperNot a substantial claim
  • Findingown data

    Models built from words and phrases as well as those from automatically generated topics achieve significantly higher out-of-sample prediction accuracies than a standard lexica for each variable of interest (gender, age, and personality).

    “Models built from words and phrases as well as those from automatically generated topics achieve significantly higher out-of-sample prediction accuracies than a standard lexica for each variable of interest (gender, age, and personality).”
    Quote in the paperSubstantial claim
  • Findingown data

    Males use the possessive 'my' when mentioning their 'wife' or 'girlfriend' more often than females use 'my' with 'husband' or 'boyfriend'.

    “males use the possessive ‘my’ when mentioning their ‘wife’ or ‘girlfriend’ more often than females use ‘my’ with ‘husband’ or 'boyfriend’”
    • Also stated in: § Contributions · boosted
      “males prefer to precede ‘girlfriend’ or ‘wife’ with the possessive ‘my’ significantly more than females do for ‘boyfriend’ or ‘husband’.”
    Quote in the paperSubstantial claim
  • Findingown data

    Introverts show an interest in Japanese media (such as anime, pokemon, manga and Japanese emoticons: ˆ_ˆ).

    “introverts show an interest in Japanese media (such as anime, pokemon, manga and Japanese emoticons: ˆ_ˆ).”
    Quote in the paperSubstantial claim
  • Findingown data

    The prediction model for gender yielded state-of-the-art results for predictive models based entirely on language, yielding an out-of-sample accuracy of 91.9%.

    “Additionally, our prediction model for gender yielded state-of-the-art results for predictive models based entirely on language, yielding an out-of-sample accuracy of 91.9%.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 73
  • May combine two claims. 44
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Subjects living in high elevations talk about the mountains.

    “subjects living in high elevations talk about the mountains”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The statuses ranged over 34 months, from January 2009 through October 2011.

    “The statuses ranged over 34 months, from January 2009 through October 2011.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Rerunning the gender analysis on reduced random samples resulted in 0 significant correlations for 50 subjects, 7 for 500, 1,489 for 5,000, and 13,152 for 50,000.

    “Rerunning our language of gender analysis on reduced random samples of our subjects resulted in the following number of significant correlations (Benjamini-Hochberg tested ): 50 subjects: 0 significant correlations, 500 subjects: 7 correlations; 5,000 subjects: 1,489 correlations; 50,000 subjects: 13,152 correlations”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    When correlating geographic elevation with language features adjusted for gender and age, 'beach' is the most distinguishing feature for low elevation localities, and 'the mountains' is among the most distinguishing features for high elevation localities.

    “when we correlate the target variables geographic elevation with language features (, , adjusted for gender and age), we find ‘beach’ the most distinguishing feature for low elevation localities, and ‘the mountains’ to be among the most distinguishing features for high elevation localities”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Articles are highly predictive of males, being older, and openness.

    “Articles are highly predictive of males, being older, and openness.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    Openness had the least agreement with the comparison study; roughly half of the results were in the opposite direction from prior work.

    “Openness had the least agreement with the comparison study; roughly half of our results were in the opposite direction from the prior work.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    A drunk topic appears for 19 to 22 year olds and a more reserved beer topic for 23 to 29 year olds.

    “consider the drunk topic (e.g. ‘drunk’, ‘hangover’, ‘wasted’) that appears for 19 to 22 year olds and a more reserved beer topic (e.g. ‘beer’, ‘drinking’, ‘ale’) for 23 to 29 year olds.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The most distinguishing topics for low elevations are (beach, sand, sun, water, waves, ocean, surf, sea, toes, sandy, surfing, beaches, sunset, Florida, Virginia), and for high elevations are (Colorado, heading, headed, leaving, Denver, Kansas, City, Springs, Oklahoma, trip, moving, Iowa, KC, Utah, bound).

    “Similarly, we find the most distinguishing topics to be (beach, sand, sun, water, waves, ocean, surf, sea, toes, sandy, surfing, beaches, sunset, Florida, Virginia) for low elevations and (Colorado, heading, headed, leaving, Denver, Kansas, City, Springs, Oklahoma, trip, moving, Iowa, KC, Utah, bound) for high elevations.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The LDA generative model assumes that documents (i.e. Facebook messages) contain a combination of topics, and that topics are a distribution of words.

  • May combine two claims.

    Findingown data

    A school related topic appears for 13 to 18 year olds, while a college related topic appears for 19 to 22 year olds.

    “we see a school related topic for 13 to 18 year olds (e.g. ‘school’, ‘homework’, ‘ugh’), while we see a college related topic for 19 to 22 year olds (e.g. ‘semester’, ‘college’, ‘register’).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The anger category is highly predictive for males, younger individuals, those low in agreeableness and conscientiousness, and high in neuroticism.

    “As a content-related language variable, the anger category also proved highly predictive for males as well as younger individuals, those low in agreeableness and conscientiousness, and high in neuroticism.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Each person in the sample wrote an average of 4,129 words over 206 status updates, and 20 words per update.

    “From this sample each person wrote an average of 4,129 words over 206 status updates, and thus 20 words per update.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In the gender results, 'my wife' and 'my girlfriend' emerged as strongly correlated in males, while 'husband' and 'boyfriend' were most predictive for females.

    “we noticed ‘my wife’ and ‘my girlfriend’ emerged as strongly correlated in the male results, while simply ‘husband’ and ‘boyfriend’ were most predictive for females.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The discarded relationships are still statistically significant.

    “These discarded relationships are still statistically significant, but removing them provides more room in the visualizations for other significant results, making the visualization as a whole more meaningful.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Personality distributions in the sample are quite typical with means near zero and standard deviations near 1.

    “Personality distributions are quite typical with means near zero and standard deviations near 1.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The restricted sample wrote a total of 309 million words across 15.4 million status updates.

    “This resulted in volunteers, writing a total of 309 million words (700 million feature instances of words, phrases, and topics) across 15.4 million status updates.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Females were more likely to precede 'husband' or 'boyfriend' with 'her' or 'amazing' and a greater variety of words.

    “females were more likely to precede ‘husband’ or ‘boyfriend’ with ‘her’ or ‘amazing’ and a greater variety of words, which is why ‘my husband’ was not more predictive than ‘husband’ alone.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    LIWC does not contain a party topic, which emerges as a key distinguishing feature for extraverts.

    “although LIWC has socially related categories, it does not contain a party topic, which emerges as a key distinguishing feature for extraverts.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The LDA generative model assumes that documents (i.e. Facebook messages) contain a combination of topics, and that topics are a distribution of words.

  • May combine two claims.

    Findingown data

    Using only features selected via DLA (significantly correlated, Bonferroni-corrected) rather than all features, open-vocabulary features again significantly improve over LIWC, with no significant changes in overall accuracy.

    “We again found significant improvement from using the open-vocabulary features over LIWC and no significant changes in accuracy overall.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    We extracted words, phrases, and topics from millions of Facebook messages and found the language that correlates most with gender, age, and five factors of personality.

    “We extracted words, phrases, and topics (automatically clustered sets of words) from millions of Facebook messages and found the language that correlates most with gender, age, and five factors of personality.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    'We' increases approximately linearly after age 22, whereas 'I' monotonically decreases.

    “Figure 5C shows ‘we’ increases approximately linearly after the age of 22, whereas ‘I’ monotonically decreases.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Our current results give detailed insights (e.g., males prefer to precede 'wife' with the possessive 'my' more so than females precede 'husband' with 'my').

    “give detailed insights (males prefer to precede ‘wife’ with the possessive ‘my’ more so than females precede ‘husband’ with ‘my’).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Our current results tie in with other research (e.g., neurotic people disproportionately use the phrase 'depressed').

    “tie in with other research (neurotic people disproportionately use the phrase ‘depressed’)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Extraverts were more likely to mention social words such as 'party', 'love you', 'boys', and 'ladies', whereas introverts were more likely to mention words related to solitary activities such as 'computer', 'Internet', and 'reading'.

    “extraverts were more likely to mention social words such as ‘party’, ‘love you’, ‘boys’, and ‘ladies’, whereas introverts were more likely to mention words related to solitary activities such as ‘computer’, ‘Internet’, and ‘reading’.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Argamon et al. used factor analysis and identified 20 coherent components of word use to link gender and age, showing male components of language increase with age while female factors decrease.

    “Similar to our finding of 2000 topics (clusters of semantically-related words), Argamon et al. used factor analysis and identified 20 coherent components of word use to link gender and age, showing male components of language increase with age while female factors decrease [32].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32] Argamon S, Koppel M, Pennebaker J, Schler J (2007) Mining the blogosphere: age, gender, and the varieties of self-expression. First Monday 12. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Volunteers agreed to written informed consent.

    “Volunteers agreed to written informed consent.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Gilbert finds words and phrases that distinguish communication up or down a power-hierarchy across 2044 Enron emails.

    “Lastly, Gilbert finds words and phrases that distinguish communication up or down a power-hierarchy across 2044 Enron emails [45].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [45] Gilbert E (2012) Phrases that signal workplace hierarchy. In: Proceedings of the ACM 2012 conference on Computer Supported Cooperative Work. ACM, pp. 1037–1046. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Mehl et al. found neurotic and agreeable people tend to use more first-person singulars, people low in openness talk more about social processes, extraverts use longer words.

    “They found similar results to Pennebaker & King and that neurotic and agreeable people tend to use more first-person singulars, people low in openness talk more about social processes, extraverts use longer words.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] M Mehl, 2006, Personality in its natural habitat: manifestations and implicit folk theories of personality in daily life, Journal of personality and social psychology source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationown data

    The study's results tie in with other research.

    “tie in with other research (e.g., neurotic people disproportionately use the phrase ‘sick of’ and the word ‘depressed’)”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Males use more formal, affirmation, and informational words, while females use more social interaction, and deictic language.

    “Other works have found males use more formal, affirmation, and informational words, while females use more social interaction, and deictic language [33]–[36].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [33]–[36]

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Iacobelli et al. found open-vocabulary features, such as bigrams, to be better predictors than LIWC features.

    “Not limited to categorical language they found open-vocabulary features, such as bigrams, to be better predictors than LIWC features.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [67] Iacobelli F, Gill AJ, Nowson S, Oberlander J (2011) Large scale personality classification of bloggers. In: Proc of the 4th int conf on Affect comput and intel interaction. Springer-Verlag, pp. 568–577. source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    The study's data-driven collection of words, phrases, and topics yields a comprehensive description of differences between groups for any given attribute, and allows one to find unexpected results.

    “This yields a comprehensive description of the differences between groups of people for any given attribute, and allows one to find unexpected results.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationown data

    The study's analyses shed new light on psychosocial processes.

    “Our analyses shed new light on psychosocial processes yielding results that are face valid”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Yarkoni identified single word results which would not have been caught with LIWC, such as 'hug' correlating positively with agreeableness.

    “He identified single word results which would not have been caught with LIWC, such as ‘hug’ correlating positively with agreeableness (there is no physical affection category inLIWC), but, considering the sparse nature of words, 406 blogs does not result in comprehensive view.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [27] T Yarkoni, 2010, Personality in 100,000 Words: A large-scale analysis of personality and word use among bloggers, Journal of Research in Personality source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The 2007 version of LIWC includes 64 different categories of language.

    “The 2007 version of LIWC includes 64 different categories of language ranging from part-of-speech (i.e. articles, prepositions, past-tense verbs, numbers,...) to topical categories (i.e. family, cognitive mechanisms, affect, occupation, body,...), as well as a few other attributes such as total number of words used [11].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [11] JW Pennebaker, 2007, The development and psychometric properties of liwc2007 the university of texas at austin, LIWCNET

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Including other variables in the regression allows adjustment for covariates such as gender and age to provide the unique effect of a given language feature on each psychosocial variable.

    “Including other variables allows us to adjust for covariates such as gender and age to provide the unique effect of a given language feature on each psychosocial variable.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Dictionary methods may signal something different when used in a new domain.

    “they note that dictionary methods (such as the closed-vocabulary analyses discussed here) may signal something different when used in a new domain (for example ‘crude’ may be a negative word in student essays, but be neutral in energy industry reports: ‘crude oil’).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [43] Grimmer J, Stewart BM (2013) Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    DLA is an open-vocabulary approach that automatically determines linguistic features including words, phrases, and topics from texts, rather than being limited to predefined word lists.

    “Open-vocabulary – it is not limited to predefined word lists. Rather, linguistic features including words, phrases, and topics (sets of semantically related words) are automatically determined from the texts.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    Open-vocabulary analyses can yield additional insights and more information (as measured through predictive accuracy) than traditional a priori word-category approaches.

    “Then, we show that open-vocabulary analyses can yield additional insights (correlations between personality and behavior as manifest through language) and more information (as measured through predictive accuracy) than traditional a priori word-category approaches.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    DLA finds key linguistic features that distinguish psychological and demographic attributes using stringent significance tests.

    “Discriminating – it finds key linguistic features that distinguish psychological and demographic attributes, using stringent significance tests.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    The study explores language in the broad context of Facebook, and does not claim results would hold under other smaller or larger contexts.

    “In our case we explore language in the broad context of Facebook, and do not claim our results would up under other smaller or larger contexts.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    The study's correlations were adjusted for gender and age.

    “our : Standardized multivariate regression coefficients adjusted for gender and age for this current study over Facebook”
    • Also stated in: § Contributions · plain
      “correlations adjusted for age; Bonferroni-corrected”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Females use more first-person singular pronouns, males use more articles, and older individuals use more plural pronouns and future tense verbs.

    “finding females use more first-person singular pronouns, males use more articles, and that older individuals use more plural pronouns and future tense verbs [30]–[32].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30]–[32]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    All research procedures were approved by the University of Pennsylvania Institutional Review Board.

    “All research procedures were approved by the University of Pennsylvania Institutional Review Board.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous research has suggested populations for online studies and Facebook are quite representative.

    “Furthermore, previous research has suggested populations for online studies and Facebook are quite representative [23], [24].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [23] S Gosling, 2004, Should we trust web-based studies? a comparative analysis of six preconceptions about internet questionnaires, American Psychologist source

      Not checked in the free run.

    • cites [24] M Back, 2010, Facebook profiles reect actual personality, not self-idealization, Psychological Science source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Sumner et al. examined the language of 537 Facebook users with LIWC while Holtgraves studied the text messages of 46 students. Findings from these studies largely confirmed past links with LIWC but also introduced some new links such as neurotics using more acronyms or those high in openness using more quotations.

    “Sumner et al. examined the language of 537 Facebook users with LIWC [25] while Holtgraves studied the text messages of 46 students [26]. Findings from these studies largely confirmed past links with LIWC but also introduced some new links such as neurotics using more acronyms [26] or those high in openness using more quotations [25].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [25] Sumner C, Byers A, Shearing M (2011) Determining personality traits &amp; privacy concerns from facebook activity. In: Black Hat Briefings. 1–29.

      Not checked in the free run.

    • cites [26] T Holtgraves, 2011, Text messaging, personality, and the social context, Journal of Research in Personality source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Multiple studies report that emoticons are used more often by females, but Huffaker & Calvert found males use them more in a sample of 100 teenage bloggers.

    “For example, multiple studies report that emoticons (i.e. ‘:)’ ‘:-(‘) are used more often by females [34], [36], [38], but Huffaker & Calvert found males use them more in a sample of 100 teenage bloggers [39].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [34] M Newman, 2008, Gender differences in language use: An analysis of 14,000 text samples, Discourse Processes source

      Not checked in the free run.

    • cites [36] Rao D, Yarowsky D, Shreevats A, Gupta M (2010) Classifying latent user attributes in twitter. In: Proceedings of the 2nd international workshop on Search and mining user-generated contents. ACM, pp. 37–44. source

      Not checked in the free run.

    • cites [38] Burger J, Henderson J, Kim G, Zarrella G (2011) Discriminating gender on twitter. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, pp. 1301–1309.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    Most works for gaining language-based insights in psychology are closed-vocabulary, while many works in computational linguistics are open-vocabulary, they rarely focus on insight.

    “Most works for gaining language-based insights in psychology are closed-vocabulary (for examples, see previous section), and while many works in computational linguistics are open-vocabulary, they rarely focus on insight.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Eisenstein et al. presented a sophisticated open-vocabulary language analysis of demographics.

    “Eisenstein et al. presented a sophisticated open-vocabulary language analysis of demographics [41].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41] Eisenstein J, Smith NA, Xing EP (2011) Discovering sociolinguistic associations with structured sparsity. In: Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies-Volume 1. Association for Computational Linguistics, pp. 1365–1374.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Monroe et al. explored a variety of techniques that compare two frequencies of words, and settled on a Bayesian model with regularization and shrinkage based on priors of word use.

    “For example, Monroe et al. explored a variety of techniques that compare two frequencies of words – one number for each of two groups [44]. In particular, they explored frequencies across democratic versus republican speeches and settled on a Bayesian model with regularization and shrinkage based on priors of word use.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [44] BL Monroe, 2008, Fightin&apos;words: Lexical feature selection and evaluation for identifying the content of political conict, Political Analysis source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Older individuals use more positive emotion and less negative emotion words, prefer fewer self-references, and stylistically there is less use of negation.

    “For age, the most salient findings include older individuals using more positive emotion and less negative emotion words [30], older individuals preferring fewer self-references (i.e. ‘I’, ‘me’) [30], [31], and stylistically there is less use of negation [37].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [30] J Pennebaker, 2003, Words of wisdom: Language use over the life span, Journal of personality and social psychology source

      Not checked in the free run.

    • cites [31] Chung C, Pennebaker J (2007) The psychological function of function words. Social communication: Frontiers of social psychology : 343–359.

      Not checked in the free run.

    • cites [37] Schler J, Koppel M, Argamon S, Pennebaker J (2006) Effects of age and gender on blogging. In: Proceedings of 2006 AAAI Spring Symposium on Computational Approaches for Analyzing Weblogs. pp. 199–205.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Premisecitation

    One should always be careful generalizing new results outside of the domain they were found as language is often dependent on context.

    “One should always be careful generalizing new results outside of the domain they were found as language is often dependent on context [40].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [40] P Eckert, 2008, Variation and the indexical field1, Journal of Sociolinguistics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Pennebaker & King found agreeable people using more articles, introverts and those low in conscientiousness using more words signaling distinctions, and neurotic individuals using more negative emotion words.

    “Their results were quite consistent across such domains, finding patterns such as agreeable people using more articles, introverts and those low in conscientiousness using more words signaling distinctions, and neurotic individuals using more negative emotion words.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [21] J Pennebaker, 1999, Linguistic styles: language use as an individual difference, Journal of personality and social psychology source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Openness exhibits the most variation across conditions of other studies, and its component traits are most loosely related.

    “openness exhibits the most variation across conditions of other studies (for examples, see [25], [27], [65]), and its component traits are most loosely related [85].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [25] Sumner C, Byers A, Shearing M (2011) Determining personality traits &amp; privacy concerns from facebook activity. In: Black Hat Briefings. 1–29.

      Not checked in the free run.

    • cites [27] T Yarkoni, 2010, Personality in 100,000 Words: A large-scale analysis of personality and word use among bloggers, Journal of Research in Personality source

      Not checked in the free run.

    • cites [65] Golbeck J, Robles C, Edmondson M, Turner K (2011) Predicting personality from twitter. In: Proc of the 3rd IEEE Int Conf on Soc Comput. 149–156. doi:<ext-link xmlns:xlink="http://www.w3.org/1999/xlink" ext-link-type="uri" xlink:href="http://dx.doi.org/978-0-7695-4578-3/11" xlink:type="simple">978-0-7695-4578-3/11</ext-link>.

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Open-vocabulary words, phrases, and topics can be identified using ordinary least squares regression.

    “Similar to word categories, distinguishing open-vocabulary words, phrases, and topics can be identified using ordinary least squares regression.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Including other variables as covariates gives the unique effect of the target explanatory variable.

    “we include other variables (e.g. age and gender) as covariates to get the unique effect of the target explanatory variable.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductcitation

    Word and phrase counts are normalized by each subject's total word use and Anscombe transformed for variance stabilization.

    “All word and phrase counts are normalized by each subject's total word use (), and we apply the Anscombe transformation [71] to the normalized values for variance stabilization”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [71] FJ Anscombe, 1948, The transformation of poisson, binomial and negative-binomial data, Biometrika source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    This is the first use of status updates from this dataset to link with personality.

    “Previously, profile information (i.e. network metrics, relationship status) from users in this dataset have been linked with personality [83], but this is the first use of its status updates.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [83] Bachrach Y, Kosinski M, Graepel T, Kohli P, Stillwell D (2012) Personality and patterns of facebook usage. Web Science. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    The latent variable of topics can be estimated through Gibbs sampling.

    “since the words in a document are known, the latent variable of topics can be estimated through Gibbs sampling [74].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [74] A Gelfand, 1990, Sampling-based approaches to calculating marginal densities, Journal of the American statistical association source

      Not checked in the free run.

    • The authors state this as an assumption: The LDA generative model assumes that documents (i.e. Facebook messages) contain a combination of topics, and that topics are a distribution of words.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Gender can be adjusted for in the LOESS regression model by including it as a covariate when training and using a neutral gender value when plotting.

    “We are able to adjust for gender in the regression model by including it as a covariate when training the LOESS model and then using a neutral gender value when plotting.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Study design and conductcitation

    All results in the gender analysis are significant at Bonferroni-corrected p < 0.001.

    “All results are significant at Bonferroni-corrected [76] .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [76] OJ Dunn, 1961, Multiple comparisons among means, Journal of the American Statistical Association source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The personality questionnaire has a retest reliability of [12].

    “Participants could take 20 to 100 item versions of the questionnaire, with a retest reliability of [12].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [12] Kosinski M, Stillwell D, Graepel Y (2013) Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Sciences (PNAS). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The analysis is restricted to words and phrases used by at least 1% of subjects, keeping the focus on common language.

    “Lastly, we restrict our analysis to those words and phrases which are used by at least 1% of our subjects, keeping the focus on common language.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Coefficients are considered significant if they are less than a Bonferroni-corrected two-tailed p of 0.001.

    “Since we explore many features at once, we consider coefficients significant if they are less than a Bonferroni-corrected [76] two-tailed of 0.001.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [76] OJ Dunn, 1961, Multiple comparisons among means, Journal of the American Statistical Association source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    DLA uses simple, fast, and readily accepted statistical techniques.

    “Simple – it uses simple, fast, and readily accepted statistical techniques.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The emoticon-aware tokenizer captures emoticons like '3' (a heart) or ':-)' (a smile), which most tokenizers incorrectly divide up as separate pieces of punctuation.

    “We built an emoticon-aware tokenizer on top of Pott's “happyfuntokenizer” allowing us to capture emoticons like ‘3’(a heart) or ‘:-)’ (a smile), which most tokenizers incorrectly divide up as separate pieces of punctuation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    In the word clouds, word size is scaled by correlation strength and color represents frequency, so larger words indicate stronger correlations and darker colors indicate more frequently used words.

    “we scale word size according to the strength of the correlation of the word with the demographic or psychological measurement of interest, and we use color to represent frequency over all subjects; that is, larger words indicate stronger correlations, and darker colors indicate more frequently used words.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Language features containing information already provided by a feature with higher correlation are automatically pruned to reduce redundancy.

    “To reduce the redundancy of results, we automatically prune language features containing information already provided by a feature with higher correlation.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The analysis is restricted to Facebook users who indicate English as a primary language, have written at least 1,000 words, are less than 65 years, and indicate both gender and age.

    “We restrict our analysis to those Facebook users meeting certain criteria: They must indicate English as a primary language, have written at least 1,000 words in their status updates, be less than 65 years (to avoid the non-representative sample above 65), and indicate both gender and age (for use as controls).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    DLA identifies the most distinguishing language features for any given attribute.

    “Our DLA method identifies the most distinguishing language features (words, phrases: a sequence of 1 to 3 words, or topics: a cluster of semantically related words) for any given attribute.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Correlation results are values from an ordinary least squares linear regression adjusted for gender and age to give the unique effect of the target variable.

    “In our case, correlation results are values from an ordinary least squares linear regression where we can adjust for gender and age to give the unique effect of the target variable.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Effect sizes tend to be relatively smaller as sample sizes increase and become more stable.

    “One should keep in mind that it is often found that effect sizes tend to be relatively smaller as sample sizes increase and become more stable [84].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [84] J Sterne, 2000, Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature, J Clin Epidemiol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Scaling by correlation coefficient rather than frequency gives clouds that distinguish a given outcome.

    “Further, scaling by correlation coefficient rather than frequency gives clouds that distinguish a given outcome.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Study design and conductnone given

    Keeping phrases with PMI values greater than a threshold ensures that kept phrases are informative parts of speech and not just accidental juxtapositions.

    “In practice, we kept phrases with values greater than , where is the number of words contained in the phrase, ensuring that phrases we do keep are informative parts of speech and not just accidental juxtapositions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The predictive models were built using 75% of participants as training data and 25% as test data.

    “We randomly sampled 25% of our participants as test data, and used the remaining 75% as training data to build our predictive models.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Exploratory analyses like DLA change the process from testing theories with observations to data-driven identification of new connections.

    “Exploratory analyses like DLA change the process from that of testing theories with observations to that of data-driven identification of new connections [97],[98].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [97],[98]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    A linear support vector machine is used for classifying gender, and ridge regression for predicting age and each personality factor.

    “We use a linear support vector machine (SVM) [92] for classifying the binary variable of gender, and ridge regression [93] for predicting age and each factor of personality.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [92] RE Fan, 2008, LIBLINEAR: A library for large linear classification, Journal of Machine Learning Research

      Not checked in the free run.

    • cites [93] A Hoerl, 1970, Ridge regression: Biased estimation for nonorthogonal problems, Technometrics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationown data

    We discovered insights not found previously.

    “We discovered insights not found previously”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    The increase in 'we' is a proxy for social integration, suggesting the increasing importance of friendships and relationships as people age.

    “We take this as a proxy for social integration [19], suggesting the increasing importance of friendships and relationships as people age.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [19] JW Pennebaker, 2003, Psychological aspects of natural language use: our words, our selves, Annual Review of Psychology source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    The increased performance could be attributed to the set of language features, a strong predictive algorithm (SVM), and the large sample of Facebook data.

    “Our increased performance could be attributed to our set of language features, a strong predictive algorithm (the support vector machine), and the large sample of Facebook data.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationown data

    Many results confirm previous research, demonstrating the instrument's face validity.

    “While many of our results confirm previous research, demonstrating the instrument's face validity, our word clouds also suggest new hypotheses.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationcitation

    Under different contexts, it is likely some results would differ.

    “Under different contexts, it is likely some results would differ.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductreasoning

    The predictive model is created without any outcome information outside of the training data, making the test data an out-of-sample evaluation.

    “Thus, the predictive model is created without any outcome information outside of the training data, making the test data an out-of-sample evaluation.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    DLA in social media is an unobtrusive and non-reactive window into the social and psychological characteristics of people's everyday concerns.

    “Differential language analysis (DLA) in social media is an unobtrusive and non-reactive window into the social and psychological characteristics of people's everyday concerns.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    This work is subject to sampling and social desirability biases.

    “Like most studies in the social sciences, this work is still subject to sampling and social desirability biases.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Facebook status updates are self-descriptive, personal, and have emotional content.

    “as “status” updates are self-descriptive, personal, and have emotional content [7].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [7] Kramer A (2010) An unobtrusive behavioral model of gross national happiness. In: Proc of the 28th int conf on Human factors in comp sys. ACM, pp. 287–290. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisereasoning

    Unlike surveys and questionnaires, Facebook language allows researchers to observe individuals as they freely present themselves in their own words.

    “unlike surveys and questionnaires, Facebook language allows researchers to observe individuals as they freely present themselves in their own words.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    Open-vocabulary approaches consider all words encountered and thus are able to adapt well to the evolving language in social media or other genres.

    “They consider all words encountered and thus are able to adapt well to the evolving language in social media or other genres.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Features were run through principal component analysis to reduce the feature dimension to half of the number of users.

    “Features were first run through principal component analysis to reduce the feature dimension to half of the number of users.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The regularization parameter was set by validation over the training set.

    “Both SVM classification and ridge regression utilize a regularization parameter, which we set by validation over the training set (we defined a small validation set of 10% of the training set which we tested various regularization parameters over while fitting the model to the other 90% of the training set in order to select the best parameter).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Open-vocabulary approaches save labor in creating categories.

    “Open-vocabulary approaches also save labor in creating categories.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Study design and conductown data

    Our current results have face validity.

    “Our current results have face validity (subjects in high elevations talk about ‘the mountains’)”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    Automatically clustering words into coherent topics allows one to potentially discover categories that might not have been anticipated.

    “However, automatically clustering words into coherent topics allows one to potentially discover categories that might not have been anticipated (e.g. sports teams, kinds of outdoor exercise, or Japanese cartoons).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    Language use is objective and quantifiable behavioral data.

    “Language use is objective and quantifiable behavioral data [96]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [96] Ireland ME, Mehl MR (2012) Natural language use as a marker of personality. (in press) Oxford Handbook of Language and Social Psychology.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Language connections with psychosocial variables are often dependent on context.

    “Language connections with psychosocial variables are often dependent on context [40].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [40] P Eckert, 2008, Variation and the indexical field1, Journal of Sociolinguistics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Open-vocabulary approaches are transparent in that the exact words driving correlations are not hidden behind a level of abstraction.

    “They are also transparent in that the exact words driving correlations are not hidden behind a level of abstraction.”
    Quote in the paperNot a substantial claim
95 of 95

  • The paper relies on this without support.

    Facebook messages are often personal and describe everyday concerns.

    “Besides advantages due to the size of the data, the content is often personal and describes everyday concerns.”

  • The authors state this as an assumption.

    The LDA generative model assumes that documents (i.e. Facebook messages) contain a combination of topics, and that topics are a distribution of words.

    “The LDA generative model assumes that documents (i.e. Facebook messages) contain a combination of topics, and that topics are a distribution of words”
2 of 2

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Shaped personality verifiably influences LLM behaviour in common downstream tasks, such as writing social media posts.

    “shaped personality verifiably influences LLM behaviour in common downstream (that is, subsequent) tasks, such as writing social media posts12”

    From A psychometric framework for evaluating and shaping personality traits in large language models (2025) · cites this paper as [12]

    The Discussion (page unavailable) reports correlations between personality factors and language in Facebook posts written by people. The Data Set section (page unavailable) confirms these were participants’ status updates, not LLM outputs. The paper therefore supports a link between human personality and social-media language, but not the claim that shaped personality influences LLM behaviour in downstream writing tasks. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • An emerging body of work has begun to explore the use of automated conversational agents and language models to identify unique traits in social media users' communication styles.

    “In addition, an emerging body of work has begun to explore the use of automated conversational agents and language models to identify unique traits in social media users' communication styles [93].”

    From What Do We Mean When We Talk about Trust in Social Media? A Systematic Review (2023) · cites this paper as [93]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Age cutoff conflicts with the reported age range

    Minor

    The eligibility criteria exclude participants aged 65, but the sample summary reports an age range extending to 65.

  • Conflicting counts of topic features

    Minor

    Figure 1 describes 500 topic features, while the feature-extraction section says the analysis produced 2,000 topics. The text does not explain a reduction from 2,000 to 500.

  • Different supplement figures named for the same personality results

    Minor

    The text directs readers to Figure S2 for the remaining three personality traits, but the Figure 6 caption assigns those results to Figure S8.

  • Conflicting identification of the largest previous study

    Minor

    The contributions section names study [27] as the next largest study, although the paper elsewhere calls a different, 3,000-blogger study the largest previous study. It describes study [27] as having only 406 bloggers.

4 of 4

Cites

Flagged first, then by citations

Research works printed but not linked: 4

  1. JW Pennebaker, 2007, The development and psychometric properties of liwc2007 the university of texas at austin, LIWCNET, 1, 1
  2. Y Tausczik, 2010, The psychological meaning of words: Liwc and computerized text analysis methods, Journal of Language and Social Psychology, 29, 24, 10.1177/0261927X09351676 · DOI · Paper page
  3. Argamon S, Dhawle S, Koppel M, Pennebaker JW (2005) Lexical predictors of personality type. In: Proceedings of the Joint Annual Meeting of the Interface and the Classification Society.
  4. SD Gosling, 2000, Should we trust web-based studies? a comparative analysis of six preconceptions about internet questionnaires, American Psychologist, 59, 93, 10.1037/0003-066X.59.2.93 · DOI · Paper page

Cited by

Sorted by citations

Related

Authors
Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Dziurzynski, L., Ramones, S. M., Agrawal, M., Shah, A., Kosinski, M., Stillwell, D., Seligman, M. E. P., & Ungar, L. H.
Affiliations
  • University of Pennsylvania (US)
  • University of Cambridge (GB)
Corresponding author
Schwartz, H. Andrew
ORCID
Venue
PLoS ONE
Publisher
Public Library of Science (PLoS)
Published
2013-09-25
Volume
8
Issue
9
Pages
e73791
ISSN
1932-6203 (ISSN-L)
DOI
10.1371/journal.pone.0073791
PMC
PMC3783449
OpenAlex
W2119595472
Cited by
1,700
References
97
Citations by year
2026: 61 · 2025: 77 · 2024: 98 · 2023: 120 · 2022: 142 · 2021: 184 · 2020: 227 · 2019: 195 · 2018: 187 · 2017: 161 · 2016: 110 · 2015: 85 · 2014: 46 · 2013: 6
FWCI
123.22
Type
journal article
Language
en
Open access
Licence
Locations
  • PLoS ONE (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, other-oa)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion, cc-by)
  • Minerva Access (University of Melbourne) (submittedVersion, public-domain)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Figshare (submittedVersion, cc-by)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • Apollo (cc-by)
Funders
Robert Wood Johnson Foundation
Topics
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
  • Physical Sciences › Computer Science › Artificial Intelligence › Authorship Attribution and Profiling
  • Social Sciences › Psychology › Clinical Psychology › Personality Traits and Psychology
Keywords
personality, gender differences, age differences, social media language, Facebook messages, Big Five personality traits, neuroticism, emotional stability, language and personality, possessive pronouns, psychosocial processes
MeSH terms
Age Factors, Female, Humans, Language, Male, Personality, Sex Factors, Vocabulary, Social Media

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review