Beyond Testimony: When Online Information Sharing is not Testifying, Emily Sullivan

In a game of telephone, or as we called it “whisper down the lane”, someone whispers a sentence to someone else, and then that person whispers it to yet another person, and on and on it goes until the end of the chain. One reason the game is enjoyable is that by the end of the chain the message becomes surprisingly distorted. These distortions can occur unintentionally: things get misheard or misremembered. However, intentional distortions also occur … [please read below the rest of the article].

Image credit: Kiki Sorensen via Flickr / Creative Commons

Article Citation:

Sullivan, Emily. 2019 “Beyond Testimony: When Online Information Sharing is not Testifying.” Social Epistemology Review and Reply Collective 8 (10): 20-24.

The PDF of the article gives specific page numbers.

Articles in this Special Issue:

In a game of telephone, or as we called it “whisper down the lane”, someone whispers a sentence to someone else, and then that person whispers it to yet another person, and on and on it goes until the end of the chain. One reason the game is enjoyable is that by the end of the chain the message becomes surprisingly distorted. These distortions can occur unintentionally: things get misheard or misremembered. However, intentional distortions also occur. Intentional distortions that cannot be chalked up to a reasonable misunderstanding are usually not tolerated. It is the minor and clever intentional distortions that go unnoticed and arguably contribute to the game’s entertainment value.

The game has an implicit moral: when information goes through chains of people, part of the message can get lost or distorted along the way. In the context of the game, we wouldn’t take the sharing of information as an instance of testimony, or as a reason to believe the content of what is said. However, it is not clear when and to what extent there is a breakdown of testimony and trust when information goes through chains of people in everyday contexts. It seems obvious to apply the moral of telephone to instances of gossip, but it is not at all clear what to do with more informative content traveling through chains of people.

The Social Digital Landscape

The new digital climate of social media platforms creates the potential for news to quickly travel far and wide, passing through even longer chains of people. The ease of information sharing comes with obvious epistemic benefits. People have quick access to reliable information that they might not otherwise have access to. Moreover, it seems as though the telephone effect does not arise, even though information is passing through more and more people. Social media platforms are designed in such a way that there are structural limits to the way the original message can change. Clicking the share or retweet button leaves the exact words of the original message in place. People can add commentary to the message, but the original message is still there for everyone to see.

Despite the benefits, the social digital landscape also brings epistemic downsides. If the information that travels across the platform is in some way false or misleading, it can perpetuate epistemic harm. Such epistemic harm has especially been felt with the increase of vaccine skepticism and the increasing amount of misleading political information. In order to address the problems of false and misleading information, new attention has been given to filter bubbles (Pariser 2011), echo chambers (Nguyen 2018), and information cascades (Guille et al. 2013).

The focus is not on the worries associated with the telephone effect, i.e., that information can get lost or discarded along the way. Instead, the focus is on what motivates someone to share something, the content of what is being shared (Vosoughi et al. 2018), and how platforms rank information in ways that create filter bubbles (Masterton and Olsson 2017; Miller and Record 2013). Through all of this, the epistemic norms of testimony and belief revision can stay roughly static. Perhaps there are additional testimonial considerations when we expand beyond the testimonial exchanges between two people (Sullivan et al. 2019); however, the question of when testimony simply breaks down is absent.

I want to argue that despite having the words of the original message remain intact after going through chains of people on digital social media platforms, the moral of the telephone effect is still relevant: testimony breaks down. It is not just that there can be a breakdown of trust in the testimony one receives; I want to suggest that there is a breakdown of testimony altogether—the information shared can no longer even count as testimony.

I am not interested here in exploring when information on social media platforms can serve as an epistemically adequate source of belief, important as such a question is. Instead, I am interested in whether the content shared on social media platforms counts as testimony and if not, what the implications are.

On the Nature of Testimony

There are several different accounts of the nature of testimony. Some are quite restrictive on what speech acts counts as testimony, while others are quite permissive. For example, on the more restrictive side, Coady (1992) argues that a speech act counts as testimony only when the speaker has the intention to present evidence to someone else on a topic that is known to be in dispute, when the other person is in need of evidence. On the assurance view of testimony (e.g. Moran 2005), testimony is restricted to speech acts that come with the speaker’s assurance that the statement is true, constituting an invitation for the hearer to trust the speaker. Such views highlight the intention of the speaker and the normative character of testimony where we rebuke the testifier in the instance of false testimony (Tollefsen 2009).

On the more permissive accounts of testimony, testimony is simply tellings in general (Audi 1997; Fricker 1995;  Sosa 1991). For example, Sosa defines testimony as “a statement of someone’s thoughts or beliefs, which they might direct to the world at large and to no one in particular” (1991, 219). There is no requirement that the speaker intends to convey information at all, with the question of responsibility left open.

At first glance, Sosa’s characterization of testimony seems well suited for the social media context. Focusing on the case of Twitter, people send out messages into the internet void to reach the world at large, directed at no one person in particular. Even if a Twitter user directs her message toward her followers or tags a specific person, given the structure of Twitter, the message can spread beyond her followers, and even be published in news outlets. However, despite the natural fit of permissive accounts to information sharing on social media, such accounts have a hard time distinguishing between non-informational expressions and informational expressions (Lackey 2006). After all, the purpose of recognizing an instance of testimony is that it comes with a certain kind epistemic weight and norms for belief. Mere sayings that are not informative have a different epistemic character. As an alternative, Lackey (2006, 2008) offers a disjunctive account that relaxes the requirement for there to be a speaker intention to inform, while restricting testimony to only informational expressions. On Lackey’s view, someone testifies so long as in virtue of her communicable content, she reasonably intends to convey the given information, or she is reasonably taken by others as conveying the given information.

Depending on the view of testimony we adopt, we will characterize speech acts on social media platforms differently. For our purposes here, the only commitment that matters is that a speech act must be in some way informative to count as testimony, which leaves out instances of joking, casual comments, and (im)polite responses.

Recall that the chief worry with the telephone effect is that the words of the original message can get lost or distorted after moving through chains of people. However, in the case of social media, it is my contention that it is not that the words get lost or distorted, but that the meaning and function of the message changes and evolves as it moves through chains of people.

Consider the following case. In February 2018, a Twitter user with the added caption “there’s ‘always’ a tweet” retweeted a tweet from Donald Trump that said:

If the Dow Joans ever falls more than 1000 ‘points’ in a Single Day the sitting president should be ‘loaded’ into a very big cannon and Shot into the sun at TREMENDOUS SPEED! No excuses!

The tweet went viral, passing through chains and chains of people; all the while, the words of the original message stayed the same.

The problem is that Donald Trump never tweeted such a tweet about the Dow Jones. It was fabricated by the Twitter user, meant as a joke (Beckwith 2018). There was no invitation of trust, and no belief was expressed. The tweet was not informative. The original message was not an instance of testimony. The point here isn’t that there are jokes on Twitter, but that after passing through chains of people, the same words can become an instance of testimony.

Many people who retweeted the tweet did so with the intention to inform, calling attention to what they presumed was evidence that others needed to hear regarding the President of the United States. In so doing, they transformed a piece of non-informational content into one with informational content, albeit false informational content. It was not that the words were getting distorted as it moved through chains of people: it was the very informativeness of the message that was distorted.

The informativeness of a message can change not only with regards to speech acts that start as non-informative and becomes informative. There are also speech acts that can start out as informative that after passing through chains of people cease to be informative.

What to Believe

Consider that, somewhere on Twitter, someone firmly believes that her child was injured by vaccines. Consider further that this person authors a tweet saying that vaccines are unsafe and cites a study that she believes provides scientific evidence that vaccines are unsafe. With this tweet she is intending to inform, and she is expressing a belief that she has. Imagine further that the tweet gains traction and starts to move through chains and chains of people. Some of these people also take the tweet to be informative and pass it along to inform others. However, there are also vaccine advocates on Twitter that believe anyone who would post such a tweet is idiotic and stupid. Such an advocate retweets the tweet with the added caption “wow”. This sharing of the vaccine message is not informing. Rather, the vaccine advocate is making what Lackey (2006) calls a non-informative casual remark about what he takes to be the current state of stupidity online.

The vaccine advocate is communicating by passing along the same original message: “Look at the crazy!” It is assumed that no reasonable person could take the content of the original message as a case of informing; only unreasonable people would do so. The vaccine advocate is not testifying. So while it could be possible for someone to inform others about the “current state of stupidity” with a similar tweet, that need not be the case. The point here is that on social media platforms, especially those like Twitter, the function of a speech act that contains the same original message can change and evolve as it moves through chains of people. Even the same tweet with the same added caption “wow” could on the one hand represent the type of causal remark mentioned above, but also signify informativeness coming from a vaccine skeptic, aiming to draw attention to the evidence cited—“Look at this shocking piece of evidence!” Lackey (2006) stresses that there needs to be an obvious pragmatic implication connecting the content of the message to the information (or lack of information) that is being conveyed. The problem is that as a message moves through chains of people, the pragmatic implications change, and the original implication can get lost and distorted. The epistemic climate on social media platforms, Twitter in particular, is a series of over-hearers engaging in context shifting. In this type of environment, the very concept of testimony becomes elusive and unhelpful.

The cases of information sharing on social media platforms that we are especially interested in, the spread of misleading and false information, are especially vulnerable to evolving testimonial functions. Focusing our attention on testimony ignores the complexity of how the nature of a speech act changes and evolves as it moves through the network. it has implications for how we should think of epistemic responsibility online, and the norms or virtues associated with information sharing. Moreover, there are implications for agent-based modeling and empirically informed work that investigates and draws conclusions about information spreading through different network structures or information cascades (e.g., see Zollman 2013).

Such models often assume, which I myself have also assumed in the past (Sullivan et al. 2019), that the type of communication between agents is testimonial. However, if what I have been arguing here is correct, and I strongly suspect that it is, such a simplification can greatly distort the epistemic problems associated with the spread of false and misleading information online, creating blind spots for possible solutions, and distorting the type of epistemic responsibilities we have when choosing to share something. Thus, our social epistemology needs to go beyond testimony. We need to consider what agents should do in an epistemic landscape where a single message can, through transmission through chains of people, cease to be an instance of testimony. It isn’t the same type of telephone effect that we are familiar with, but it has similar consequences: it erodes trust in the words of others.

Contact details: Emily Sullivan, Delft University of Technology,


Audi,  Robert. 1997.  “The  Place  of   Testimony  in  the  Fabric  of   Knowledge  and Justification.” American Philosophical Quarterly 34 (4) 405-22.

Beckwith, Ryan Teague. 2018. “No, President Trump Did Not Tweet About the ‘Dow Joans’” Time  Magaizine.

Coady, C. A. J. 1992. Testimony: A Philosophical Study. Oxford: Oxford University Press.

Fricker,  Elizabeth.  1995.  “Telling  and  Trusting:  Reductionism  and  Anti-Reductionism  in   the Epistemology of  Testimony.” Mind 104 (414): 393-411.

Guille, Adrien, Hakim Hacid, C. Favre, and Djamel Abdelkader Zighed. 2013. “Information Diffusion in Online Social Networks: A Survey. ACM Sigmod Record 42 (2): 17-28.

Lackey, Jennifer. 2006. “The Nature of Testimony.” Pacific Philosophical Quarterly 87 (2): 177-197.

Lackey, Jennifer. 2008. Learning from Words: Testimony as a Source of Knowledge. Oxford University Press on Demand.

Masterton, George and Erik J. Olsson. 2018. “From Impact to Importance: The Current State of the Wisdom-of-Crowds Justification of Link-Based Ranking Algorithms.” Philosophy & Technology 31 (4): 593-609.

Miller, Boaz and Isaac Record. 2013. “Justified Belief in a Digital Age: On the Epistemic Implications of Secret Internet Technologies.” Episteme 10 (2): 117-134.

Moran, Richard. 2005. “Getting Told and Being Believed.” Philosopher’s Imprint  5 (5): 1-29.

Nguyen, C. Thi. 2018. “Echo Chambers and Epistemic Bubbles.” Episteme 1-21. doi:10.1017/epi.2018.32.

Pariser, Eli. 2011. ​The Filter Bubble, What the Internet is Hiding from You​. Penguin.

Sosa,  Ernest.  1991.  Knowledge  in  Perspective:  Selected  Essays  in  Epistemology. Cambridge University Press.

Sullivan, Emily, Max Sondag, Ignaz Rutter, Wouter Meulemans, Scott Cunningham, Delft Bettina Speckmann, and Mark Alfano. 2019. “Can Real Social Epistemic Networks Deliver the Wisdom of Crowds? In Oxford Studies in Experimental Philosophy, Volume 3 edited by Tania Lombrozo, Joshua Knobe, and Shaun Nichols. Oxford University Press.

Tollefsen, Deborah P. 2009. “Wikipedia and the Epistemology of Testimony. Episteme 6 (1): 8-24.

Vosoughi, Soroush, Deb Roy, and Sinan Aral. 2018. “The Spread of True and False News  Online.” Science 359 (6380): 1146-1151.

Zollman, Kevin J. S. 2013. “Network Epistemology: Communication in Epistemic Communities.” Philosophy Compass 8 (1): 15-27.

Categories: Articles

Tags: , , , , , , ,

Leave a Reply