3.1 The Aim of Online Exchanges

Does the aim of online exchanges amount to or involve the sincere exchange of information? Though it might seem so at least for a good deal of our engagements online, a number of researchers have called into question how central an aim this is. Instead, researchers have pointed out how a good deal of online engagement has an expressive aim (Lynch, Reference Lynch2019), or alternatively aims at establishing one’s group affiliation. If true, this (by now familiar) point has far-reaching implications: whereas standard conversations aim at or centrally involve the sincere exchange of information with the goal of arriving at the truth, online ‘conversations’ whose raison d’etre is establishing one’s group affiliation need have no interesting connection to the truth. What is more, if the aim is to reinforce group cohesiveness, it would seem that participants will restrict themselves to what serves that aim, with the result that contributions that are seen to solidify one’s identity will tend to get expressions of approval from others (likes and the like). This will loom large in the next section, when I consider the epistemic dimensions of online exchanges.

3.2 Scope of Participants

Consider ordinary face-to-face conversations. Typically, the participants know who is in the conversation and who is not. In part this is because there are conventions that enable us to discern the contours of the conversational participants. Thus, there are conventional ways to initiate a conversation with another (or others): for example, by addressing oneself to them in a way indicating one wishes to initiate a conversation with them. There are conventional ways to signal that one is or remains a participant in an ongoing conversation: for example, by shaking one’s head to indicate uptake (or performing one or another action indicating that one is attending to the contributions). And there are conventional ways to conclude a conversation. What is more, in face-to-face settings, one can see all of those who are potentially participants in the conversation, enabling one to get a good deal of information regarding the participant status of each candidate.

The salient contrasts with online exchanges are many. For one thing, unless an individual contributes to the exchange (by posting, reposting, replying, liking the post, or what-have-you), there is no way to tell whether an individual in one’s social network is a silent ‘part’ of the conversation. The contours of the participants, then, are hard to discern even if we assume that they are determinate. But there are grounds for doubting whether the contours are determinate: is one who glances at a post or a thread, registering a comment or two only to move on quickly, a part of the conversation or not? We might stipulate that only those are part of the online conversation who explicitly contribute to it in one way or another. But in that case our stipulation would eliminate the analogue in face-to-face encounters of the silent participant: the one who tracks what is going on, updates the context accordingly, but does not otherwise contribute.

Consider, too, the conventions available through which to signal an interest in initiating or concluding a conversation, as well as those for signaling one is still a participant. To be sure, posting might be considered a way to signal an interest in initiating a conversation: but with whom? With everyone in one’s social network? With those on whose wall one’s post is seen? We might also think of tagging as a way of calling one’s attention to the conversation in the hope of including them; but not all tags are for that purpose, of course, so this signal is noisy. When it comes to indicating one’s status as a participant in the conversation, of course, one can contribute to it (commenting, reposting, liking, etc.). Still, as signals of participation these acts (of commenting, reposting, liking, and the like) are also noisy, and it remains true that there would seem to be no way of doing so silently. Finally, there is no natural or conventional way to signal the end of the conversation, and it often happens that old posts (left for finished by those participating at the time) are revived when a new contributor makes a very belated contribution.

All of these differences are exacerbated by the temporal dimension of online exchanges: while technology would permit one to see who among the active participants is online at any given time, in a good many exchanges one interacts only with those who are actively participating at the time. Participants then can be remote from one another not only in space but also in time: I might reply tomorrow to a comment you made today, keeping alive the conversation even though the intervening 24 hours have been non-active.

There is an all-important corollary of not being able to discern the set of participants in an online discussion: we cannot discern the context set itself, that is, the set of propositions that are being mutually assumed for the sake of the exchange itself. The nature of the difficulty here can be characterized by contrasting two extreme views as to how to go about addressing this: what I will call a minimalist and a maximalist approach. On the minimalist approach, one assumes the minimal number of participants consistent with the nature of the exchange itself. These will include all and only those who have responded to another’s post (whether by commenting or liking or reposting or what-have-you). Here, what is presupposed is only what we need to treat all of these individuals as mutually assuming in order to make sense of the exchange itself. On the maximalist approach, one regards the speaker’s entire social network as in on the exchange, and what we as theorists regard as presupposed is what we need to make sense of the possibility of anyone of these individuals participating in this exchange. Obviously, how much this includes will depend on the diversity of views within one’s social network, the salience of those views, and so forth.

I submit that neither the minimalist nor the maximalist approach to context-fixing is the right way to capture what is going on in the online “conversation,” and that the reason for this will generalize to other (less extreme) attempts to capture the set of things that are mutually presupposed in the exchange. The trouble is that neither of these options appears to set the right constraints on what is to count as mutually presupposed in context, and so both will err in including either too much or not enough in the context set itself. Generalizing from this, I suspect that any attempt at some hard-and-fast technique for discerning the context set will make errors of one of these two sorts – either including too much in the context set, or not enough – even if the approach itself is not as extreme as either the minimalist or the maximalist approaches just described.

Let me give examples from the US and the UK to illustrate these worries about the minimalist and the maximalist approach to context-fixing.

Start with the minimalist approach. Suppose you reside in the US and you are among the majority of US voters who disapprove of President Trump (56% as of July 2020Footnote ²), where the majority of your network disapproves of him as well. Still, you might have a few pro-Trump people among your network, and they might be outspoken on those occasions when you make anti-Trump comments. If they do, does their participation in the thread you’ve initiated inviolate your attempt to presuppose (for the purpose of discussion) such things as that Trump is aiming to normalize political practices and behaviors that ought not to be normalized? Well, the cost of allowing this would be to allow all of the trolls to set the terms of our discussion. But if we insist that this should not be allowed, then it follows that the mere fact that a person is participating, or is trying to participate (e.g. by making comments on one’s discussion thread), is not sufficient, by itself, to include their perspective as serving to fix the context set. To be sure, we might try to rectify this problem by de-friending trolls. But the problem is deeper than that: a troll who makes as if to participate properly throughout doesn’t count as preventing one from presupposing what one wants to presuppose merely because the troll is quiet about matters. It seems that the mere fact that one is participating in an online discussion doesn’t yet determine the role one plays in fixing the context set. Since minimalism assumes otherwise, it is not adequate.

Move to the maximalist approach. Suppose you reside in pre-Brexit UK and are a firm Remainer, where the vast majority of your online social network is in favor of Remain. (The point doesn’t depend on the details of the politics; I use them for the purpose of illustration only.) Still, you might have the occasional Brexiteer among your network. (Suppose they are usually quiet and don’t participate much, if at all, in your discussions.) Do their pro-Brexit views nevertheless help to fix what is mutually taken as presupposed for the purpose of your discussions? If so, you will not be able to have a discussion in which the Remain position is mutually taken for granted. But this seems weird: even if you don’t always expect to be able to take that for granted, surely sometimes – in some online exchanges you initiate – you want to be able to do so, and you expect to be able to do so. So it can’t be that the mere fact that you have a few Remainers among your social network prevents you from ever being able to do so. The maximalist approach can’t be right.

What I think actually happens: we construct the context on the fly. Some posts make clear the sort of audience they have in mind: one’s professional colleagues, or family members, or high school friends, or the politically like-minded, or fellow cat-lovers, or fellow members of an interest group of some kind, etc. Other posts are sufficiently general that they might be aimed at a much wider audience, where the contours of that audience are themselves not clearly conceived in advance. Over the course of the evolution of the discussion, participants construct the context set as needed to make sense of the exchange. Insofar as some are regarded as calling into question the intended presuppositions of the conversation, they are ignored or disallowed to continue to ‘hijack’ the discussion. But if this is correct, it makes clear that there will be many cases in which the state of the context set at a given time will be far from clear to the participants, even to those centrally invested in a productive discussion.

3.3 Scope of acts performed online

To introduce the problems surrounding the scope of the speech acts performed online, I will need first to present some basic elements of speech act theory. To begin, note that the verbal use (or utterance) of a sentence is not just the production of sounds; it is rather a meaningful use of speech. Thus if I utter ‘You will sit next to me’ to you, intending thereby to be expressing what that English sentence says, I have performed a meaningful act. We can designate this act – the meaningful act one performs when one produces a sentence intending to be expressing what the sentence says – as a locutionary act. In knowing which locutionary act I have performed, you (my intended audience) thereby know what I have said. Still, as my intended audience you can know what locutionary act I have performed, and so know what I have said, without knowing how to take or understand what I've said: I might have said this as a prediction (I am predicting you won't know anyone else at the party), a decision (we are making seating arrangements for the upcoming wedding), or a command (I say it to you under my breath in a threatening tone). We can use ‘illocutionary force’ to designate that feature of a speech act that pertains to how what is said is to be taken or understood by the audience. Thus the illocutionary force of a prediction (= the way the speaker intends the audience to understand her locutionary act) differs from that of a decision, which in turn differs from that of a command. And we can use ‘illocutionary act’ to designate the resulting type of acts themselves: predictions are a different type of illocutionary act than are decisions or commands. The case above makes clear that one and the same (type of) locutionary act might be associated with various distinct types of illocutionary act.

For its part, the Stalnaker-Roberts model of conversation allows for three general types of illocutionary act: assertions (proposals to add information to the stock of propositions that re mutually presupposed), questions (proposals to add a query to the stock of questions under discussion, including subquestions of questions currently on the list as well as clarificatory questions about previous moves or other questions), and directives (proposals to add an action on the list of what is to be done). Using the notions introduced above, we can say that assertions constitute a type of illocutionary act with its own distinctive illocutionary force (= assertoric force), questions constitute a type of illocutionary act with its own distinctive illocutionary force (= interrogative force), and directives constitute a type of illocutionary act with its own distinctive illocutionary force (= directive force). The challenge is that there are actions performed in online settings whose illocutionary force is not obviously any one of these; and in addition even when it is clear (more or less) that an online act is of one of these three illocutionary types, it appears to be significantly different than standard acts of that type (in face-to-face settings). I will take these up in order.

There are many acts performed in the context of online discussions of which it is not obvious that their illocutionary force is one of the three just described. Here I mention four: posting, reposting/retweeting, liking, and (hash)tagging.

Of all of the ‘speech acts’ performed online, the post is the one that might seem the easiest to incorporate into the Stalnaker-Roberts model: isn’t the act of posting simply the act of assertion itself – at least when one’s post purports to say how things are? (By this I mean to exclude posts that are clearly intended as venting, or as merely expressive in some other way, as well as posts that extend invitations, etc.). There is much to recommend this analysis regarding posts in which one purports to say how things are. Still, there are two complications that are worth highlighting, as both of these render the construal of such posts as assertions less than fully happy.

Consider the question that one finds next to one’s name when one signs on to one’s Facebook account: What’s on your mind, [name]? Interestingly, this question permits of at least two distinct readings: what I will call the expressive reading, and what I will call the topical reading. According to the expressive reading, the question asks one to give expression to one’s own state of mind – whether that involves something one is thinking about, or an emotion one is feeling, or a reaction one is having, etc. According to the topical reading, the question asks one to address oneself to a topic and say something about that topic.

Suppose that a Facebook user interprets this question in terms of the expressive reading, and that her posts are informed by this aim. Then she might take it that what she is doing is simply giving expression to her states of mind. To be sure, we can still see her as making assertions in this case. Only if this is how she intended her post, its content will pertain in the first instance to her state of mind, rather than to the topic she is addressing. To illustrate, suppose she posts ‘My friend Tom has misbehaved’, intending thereby to be expressing ‘what’s on her mind’ (as Facebook would put it). Then she will intend for her post to be understood as capturing e.g. the irritability she is presently feeling in response to something she takes Tom to have done. Even if it makes sense to regard her as having asserted as much (namely, that she is irritated etc.), such an assertion is very different from an assertion that is straightforwardly about Tom’s behavior. To see this, notice that she might be taken aback by anyone who questions her: she regards herself as having done nothing more than having expressed her own state of mind, including her own take on the world, and any attempt to question this would, in her mind, be seen as challenging her authority to say what is on her mind. This is a very different sort of activity than the one we engage in when we make assertions about the world. To be sure, her ‘take’ on the world might be called into question (as in: Tom didn’t do what she took him to have done); but if she is pressed with such an objection, she can always resort to the response, ‘Well, this is how things struck me, and that’s all I was saying in my post’.

I mention this not to defend this sort of maneuver, but rather to point to the possibility of some unclarity as to what, precisely, one is doing when one posts on Facebook. My claim is that this unclarity remains even if we restrict ourselves to posts in which one purports to say how things are. And so, even after we agree that posts are assertions in the Stalnaker-Roberts’ sense, this possible unclarity makes it unclear what it is that is being asserted in any given case.

There is one other aspect of posting that makes it somewhat hard to accommodate it within the Stalnaker-Roberts model. Whereas that model aims to capture face-to-face exchanges between conversational participants, posts can have the feel of public announcements rather than contributions to a conversation itself. That is, one who posts is doing something more like broadcasting to a wide (indeterminate) audience, than talking to a determinate set of individuals. If this is right, of course, then the whole Stalnaker-Roberts model is not applicable in the first place – but then again, neither would it be correct to say that we engage in conversation online. I do not raise this to endorse the idea that we do not have conversations online. On the contrary, I do think we engage in (something very much like) conversations online; my present point is only that they have very different features than those of face-to-face conversations. Seen in this light, posts can have the feel like public announcements more than they are claims made in the give-and-take of a speech exchange is one such feature.

But posts are not the only ‘speech acts’ online that are hard to fit into the Stalnaker-Roberts model. Consider next the repost or retweet, when another person’s post, as such, is ‘forwarded’ in one’s name. These are often (typically?) interpreted as a re-affirmation the content of the original post or tweet. And if things were this simple, reposts/retweets could be seen as a kind of assertion in which one re-asserts a content previously asserted by another – presumably with the point of extending the dissemination of that information to one’s own social network. The difficulty is that not all retweets are endorsements. There are many motives for retweeting or reposting a previously-made post. At its most generic level, the rationale is that of bringing something to the attention of one’s social network; but one can have all sorts of reasons for wanting to do this, not all of which include endorsement. (Perhaps it will be obvious to the most salient members of one’s online social network that one regards the original post with contempt, or irony; perhaps the point is some sort of ‘in’-joke among one’s online social network; and so forth.) And even when one does endorse the content, sometimes the point of reposting/retweeting is not to re-assert what was asserted previously, but simply to register or signal one’s own endorsement of it. (Such an act puts the focus on one’s own attitude toward the content posted, rather than on the alleged truth of that post.) No doubt, the difficulty in interpreting a retweet or repost is related to the difficulty of discerning the contours of the conversation (who is included, and who is not), as well as the corresponding difficulty of discerning what is being presupposed in the context. But even if the difficulty of discerning the context is more fundamental, still, the challenge of assigning an illocutionary force to a retweet/repost, and even determining whether that force is one of the three main types postulated by the Stalnaker-Roberts model, remains.

Next, consider the act of ‘liking’ another’s post. This act is even harder to interpret than is the act of retweeting or reposting. What we might call its ‘pragmatic significance’ – what it intends to convey regarding how it is to be taken by others in the conversation – can be any of the following: I endorse what you’ve posted; I like what you’ve posted; I support you in having posted this; I like you; I have read what you posted with interest; you are on my mind as you post this; I am following this thread with interest; and many others besides. To be sure, the set of possibilities here may be narrowed e.g. on Facebook, where one has other options: in addition to an emoji for ‘like’, there are emojis for ‘love’, ‘care’, ‘ha ha’, ‘wow’, ‘sad’, and ‘angry’ (the terms are from Facebook). And when the context makes things clear (but see below), interpretations may be narrowed further. Even so, there are a great many occasions on which it is hard to see how to interpret a mere ‘like’, and in any case it is far from clear how to fit this action into the tripartite list of actions postulated by the Stalnaker-Roberts model. These acts seem more expressive than any of the three acts postulated by that model: far from proposing to add information, or a question, or a task-to-be-done, they merely express one’s attitude. No doubt, in this way they can be seen as adding the information that one expressed such an attitude to the common ground; but they are not thereby to be represented as an assertion. (As Stalnaker himself noted, all sorts of things add information to the common ground without counting as assertions: any salient public act one performs will do just this, as will salient events not involving any agent.)

Finally, consider the act of tagging. Here I have in mind tagging on both Facebook (an act involving the use of the person’s Facebook name) or on Twitter (an act involving the use of the ‘@’ sign followed by their Twitter handle). (Note that both of these are distinct from the use of the hashtag used on Twitter.) The act of tagging someone on a post can be performed with any of a variety of distinct intentions in doing so: to get the target’s attention; to indicate to the target that s/he is being discussed; to indicate to the target that s/he is being thought about in connection with the post; to elicit from the target some response; and so forth. Assuming that the relevant intention is discerned by the audience (including but not limited to the target), once again, we might think that the statement that the speaker has the relevant intention is added to the common ground; but again this does not make the act one of assertion. The act would seem more like that of addressing oneself to someone than it would an assertion, though on occasion, when intended to elicit a response from the target, it might be construed as an instruction (to the target to respond). But since this will not cover all cases of tagging, the act of tagging itself is not of a type that should be identified with any of the three types of act postulated by the Stalnaker-Roberts model. (This is not particularly surprising; the point of the act of tagging is not to add any information or question or instruction to the common ground, so much as to capture another’s attention for some purpose or other.)

While I will not have much to say about it, the act of hashtagging on Twitter is distinct from the act of tagging (whether on Facebook or Twitter). Though it can be directed at a single individual (whether explicitly, as in hashtagging them by name, or in some other way), the use of the hashtag is typically not so aimed. Instead, hashtagging is aimed at attracting the attention of the widest audience possible. Still, the intentions behind a hashtag can be varied. By their nature, hashtags allow followers of a topic to follow that topic, and might be seen as having that overarching aim. Still one might have various motives (with a greater or lesser degree of openness) for doing so: eliciting a response, capturing attention, adding to a conversation, and so forth. Once again, it is unsurprising that this act is not among the three postulated by the Stalnaker-Roberts model.