Account Verification's Effect on Tweet Credibility

CHI 2019 Paper CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK Does Being Verified Make You More Credible? Account Verification’s Efect on Tweet Credibility Tavish Vaidya Daniel Votipka Georgetown University University of Maryland Michelle L. Mazurek Micah Sherr University of Maryland Georgetown University ABSTRACT Verifcation’s Efect on Tweet Credibility. In CHI Conference on Many popular social networking and microblogging sites Human Factors in Computing Systems Proceedings (CHI 2019), May 4– support verifed accounts—user accounts that are deemed of 9, 2019, Glasgow, Scotland, UK. ACM, New York, NY, USA, 13 pages. public interest and whose owners have been authenticated https://doi.org/10.1145/3290605.3300755 by the site. Importantly, the content of messages contributed by verifed account owners is not verifed. Such messages 1 INTRODUCTION may be factually correct, or not. On many social networking and microblogging sites, it is This paper investigates whether users confuse authenticity no longer the case that nobody knows you’re a dog [43]. with credibility by posing the question: Are users more likely Twitter, Instagram, Reddit, Facebook, and other popular sites to believe content from verifed accounts than from non-verifed support verifed accounts—accounts whose owners have been accounts? We conduct two online studies, a year apart, with authenticated by the site [48].1 Verifed accounts are often 748 and 2041 participants respectively, to assess how the reserved for well-known organizations and individuals, and presence or absence of verifed account indicators infuences are indicated with a badge (e.g., ) that appears next to the users’ perceptions of tweets. Surprisingly, across both stud- account holder’s name (see Figure 1a). Verifed accounts ies, we fnd that—in the context of unfamiliar accounts—most are intended to allow users to easily distinguish between users can efectively distinguish between authenticity and authentic accounts of public interest and those belonging to credibility. The presence or absence of an authenticity indi- parody accounts or impostors. cator has no signifcant efect on willingness to share a tweet Importantly, accounts earn verifed status by being authen- or take action based on its contents. ticated (i.e., the user is who she claims to be). Sites specifcally do not assess an account owner’s trustworthiness when as- CCS CONCEPTS signing verifed status, and most major online community • Information systems → Social networks; Trust; • Se- sites do not require that verifed account owners post only curity and privacy → Authentication; Human and societal factual content. More plainly, posts from a verifed account aspects of security and privacy. may be factual, or not [11]. This paper examines whether users confate authenticity KEYWORDS with credibility, in the context of microblogging sites such as Social; Twitter; Credibility; Trust; Authentication Twitter. That is, are users more likely to believe, and there- ACM Reference Format: fore be willing to act upon, messages that originate from Tavish Vaidya, Daniel Votipka, Michelle L. Mazurek, and Micah verifed accounts? We defne credibility as the “believability” Sherr. 2019. Does Being Verifed Make You More Credible? Account of a piece of information—that is, the degree to which it infuences whether an individual is persuaded to adopt the Permission to make digital or hard copies of all or part of this work for opinion ofered and take action accordingly [24]. Indicators personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that of authenticity (such as a verifed account badge) signal that copies bear this notice and the full citation on the frst page. Copyrights a message originated from its purported source. Credible for components of this work owned by others than the author(s) must information may of course stem from both verifed and un- be honored. Abstracting with credit is permitted. To copy otherwise, or verifed accounts; conversely, the same is true for incorrect republish, to post on servers or to redistribute to lists, requires prior specifc information. Equating credibility with authenticity is thus a permission and/or a fee. Request permissions from [email protected]. CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK logical fallacy. © 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM. 1The terminology can vary slightly among sites; for example, Facebook ACM ISBN 978-1-4503-5970-2/19/05. $15.00 uses the term “verifed profle.” For consistency, we refer to accounts whose https://doi.org/10.1145/3290605.3300755 authenticity has been confrmed by site operators as verifed accounts. Paper 525 Page 1 CHI 2019 Paper CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK Three signifcant factors suggest that verifed badges could verifed account owners (and their posted content) should infuence people’s perceptions of user-generated content. Lit- be considered trustworthy.2 erature from psychology and other social sciences shows that The implications of users confating authenticity with cred- perceptions of credibility are more infuenced by the source ibility are worrisome. With the growing fraction of people of information than by its credulity [6, 14, 17, 21, 45]. No- who use online social media (with user-generated content) as tably, Hovland et al. found that the credibility of a message their primary source of news [2] and the recent proliferation is infuenced by the source, setting, and presentation of ar- of fake news [54], a possible inability to judge the credibil- guments of the message along with receivers’ predisposition ity of information from (un)verifed accounts is especially toward the message [24, pg. 10-13]. Berlo et al. found that the disconcerting. message’s source afects message credibility along three axes: To understand the efect of account verifcation on the safety, or whether the recipient believes the source might credibility of user-generated content, we conducted two stud- have an agenda or a reason to mislead; qualifcation, or how ies on Amazon’s Mechanical Turk comparing participants’ qualifed the source is to comment on the given topic; and dy- perceived credibility of tweets from verifed and unverifed namism, or how charismatic and persuasive the author is [3]. accounts. In both the studies, we varied factors such as the If information is widely endorsed (e.g., popularly “liked” on subject matter of tweets, the advertised credentials of the Twitter) and/or originates from a well-known source (e.g., poster, the positivity/negativity of tweets towards their sub- a verifed account), then the prior work suggests users will ject matter, and the indicators used to signify that an account perceive it as more trustworthy (i.e., safe) and the author as was verifed (e.g., a icon vs. a textual “Verifed account” more qualifed, leading to higher perceived credibility. label). We measured perceived credibility both implicitly, by Second, the computer security community has repeatedly asking about actionability, and explicitly, by asking directly shown that users have difculty in understanding the con- about the efect of the authenticity indicator. cept of authenticity [1, 8, 13, 46] and are generally confused Surprisingly, the results of Study 1 (n1=748, conducted in by security indicators [20, 40]. In particular, users do not Aug-Sep 2017) showed no evidence of correlation between easily distinguish between authenticity and integrity. For verifed account status (i.e., verifed or unverifed) and the example, users have been shown to believe the content of perceived credibility of tweets. To validate the null result, we websites so long as browser indicators (e.g., the golden lock) conducted a second study (Study 2, n2=2041, Aug-Sept 2018) show that the website is secure [13]. This suggests that users that confrmed our initial fndings from Study 1. In this paper, might mistake verifed account badges as an external vali- for brevity, we primarily discuss the second study. (Results dation that the posted content has been fact-checked. This from the frst study are provided as supplementary mate- efect could potentially be boosted by the iconography of rial.) As detailed in the next section, these fndings directly verifed account badges (e.g., Twitter’s and Facebook’s ). contradict results from prior work [35, 51]. Finally, the manner in which accounts become verifed has led to confusion. In particular, Twitter verifes accounts only 2 RELATED WORK when it deems the account owner to be of public interest [48]. Researchers have long sought to understand the factors that This has translated into a perception that authenticity indi- infuence users’ perceptions of online content. Wineburg et cators are markers of prestige and importance, not merely al. [53] survey students to assess whether they can correctly authenticity. Twitter has been criticized for granting verifed judge the credibility of information from online sources. account status to racists and other extremists [52]. (Here, the Fogg et al. [15] conduct a notable, large-scale study in which criticism is not that these extremists are not who they say more than 2600 participants reviewed websites for credi- they are, but rather that their accounts have earned “elite” bility. They fnd that the “design look” (i.e., user interface) status.) Twitter responded to this criticism by noting that of a site is the most frequent (and dominating) factor in “[v]erifcation was meant

Load more