Communicating online can be thrilling. We can interact instantly with people anywhere in the world. We can find and consult with experts on any conceivable topic. And we can find kindred spirits and comrades with whom to build valuable communities not possible in offline life.
And yet, it can also be maddening. We may struggle to express our true feelings with text alone (however many emoji we throw in). We may find ourselves attacked by hateful speakers emboldened by the cover of anonymity. We may, on occasion, undergo the profoundly lonely, soul-destroying realization that we are talking to a bot and not a person. And, more commonly than all of these, we can find ourselves mired in misunderstandings and miscommunications, unclear of whom we’re talking to, what we’re talking about, and who is listening.
Experiences of the latter kind bring into relief some of the starkest differences between online and offline communication. They also make clear the limitations of traditional philosophical models of conversation. Historically, such models have assumed the following about a standard conversation. Firstly, that it’s common knowledge who is taking part in it. Usually, we can see or hear everyone present. Secondly, that it’s clear to all what the norms of the conversation are—what we can and can’t do, and what, as a group, we are oriented towards (what Craige Roberts calls the “question under discussion”). Thirdly, that there is agreement about both what has happened so far in the conversation and what is in the “common ground,” as Robert Stalnaker calls it.
Famously, David Lewis helps us understand conversations of this kind using a baseball analogy. Participants in a conversation are like players in a baseball game—we are governed by a set of shared rules, known to all; we are oriented to a shared purpose; and every move we make is logged in a “scoreboard”—only, in conversation, that scoreboard is a shared construct and not a physical object illuminated on a field. Lewis also points out that conversations are more fluid than baseball games, and so conversation participants, unlike baseball players, can change the rules of the game as they go. If I say, “Look at that!” and point at the dog running towards us, I succeed in establishing a temporary norm that that refers to the dog in front of us. If you respond with, “No, look at that!” and point to a bear chasing that dog, you change the norm immediately such that that now refers to the bear. While this model is very helpful for thinking about offline conversation, it has proven woefully inadequate when applied to online conversation.
Imagine that you decide to enter the inferno that is Elon Musk’s X and make a post. In that post, you offer some feeble political commentary. In the U.K., the new pretender for prime minister is Andy Burnham, and he just won an important by-election. The UK is also experiencing a heatwave. You write, “Burn, baby, burn!”
Your goal is to amuse the handful of friends who follow you on X and to signal your support for Burnham, and maybe enjoy the vainglory of some likes. To your mind, the contours of the conversation are clear—it’s a conversation between you and a handful of friends with whom you have established a set of conversational norms and a shared taste in bad puns.
Yet the contours in such conversations are in fact never clear. Assuming your post is public, it may be viewed not only by your friends but also by family members. Your grandfather might not understand the post and might demand that you explain it. Your boss might chastise you for making ideological comments in ways that reflect negatively on your organization. You might find yourself caught between multiple sets of norms, struggling to be a banterous friend, a patient grandchild, and a dutiful employee all at once.
The post might quickly move beyond your immediate social circles. Greenpeace might see you as enemy number one and invite the wrath of millions by sharing the post with an accompanying “How dare you make light of global warming?” The fossil fuel industry, meanwhile, may ordain you as their new mascot. And somewhere across the pond, an American might assume you’re inciting book burning and start an argument with you about the First Amendment.
Clearly, your conversation is not at all like Lewis’s baseball game, even once we admit, as he does, that conversational norms are more protean than the rules of baseball. There is no fixed set of players, no fixed set of rules, and no determinate scoreboard.
In “Context Collapse Online,” I argue that Lewis’s framework must be adapted to accommodate online communication. Online communication, I argue, is not like participating in a single baseball game. Instead, it is like attempting to play a game of baseball in a busy park in which multiple overlapping games of other kinds—rounders, cricket, softball—are happening at the same time. Keeping track of what game you’re in and who is playing it can prove impossible. This is made even more complicated by the presence of opportunists and bad faith actors who wander from game to game, attempting variously to either participate sincerely in them or to confuse and upset them. It might be unclear to others what game you are playing, and you might get blamed for violating rules of cricket when you thought you were playing baseball. You might become an unwitting hero among rounders players, though you didn’t think you were playing that game either. You might even start trying to play baseball in a way that appeals to those who think you are playing rounders.
Online communication, I propose, is best thought of as involving multiple overlapping language games with fluid boundaries and constantly shifting participants, many of whom are only partially visible to one another. Embracing this model not only helps us better appreciate the differences between online and offline communication but also helps us explain various problems we encounter online.
Clearly, the complexities of online communication make miscommunication and misunderstanding far more likely. When we find ourselves in evolving, overlapping conversations with indeterminate numbers of participants, some of whom are unknown, it can be much harder to figure out which scoreboard to refer to, and indeed, what is on that scoreboard. Without this background we may struggle to recover speakers’ intended meanings and find ourselves in disagreements about what has been said.
Moreover, this environment is ripe for hijacking, derailing, and manipulation. When multiple language games overlap, people can remove an utterance from one game and insert it into another, and in doing so, expose the speaker to unwarranted criticism or use their words in ways contrary to their intentions. The complexities of online communication, combined with its sheer scale, make pile ons much more likely, because the more conversational norms in play, the more ways there are for someone’s words to be interpreted as violating one of them.
Hate speakers can appeal to the complexities of online communication to cover their backs—“You were taking me out of context!,” they might proclaim. They also have incentives to go underground with their hate speech, designing it such that it flies under the radar in multiple different conversations. This can lead (and has led) to the creation of all kinds of pernicious dogwhistles.
Where do we go from here? One response might be to stop using social media for sincere conversation. Recent research in the U.K. suggests that fewer people are posting on social media—currently 49% report doing so, down from 61% in 2024. Yet engagement with social media is still high—it is just increasingly taking a more passive form, whereby fewer people share their own content, but many continue to consume content produced by others. Some of us, it seems, have chosen to hang up our kits and take seats in the stands instead. This likely creates its own problems. For one thing, hate speech may now go unchallenged, when in the past it would have been resisted. For another, online discourse may become less representative, which could give rise to spirals of silence and the impression that extreme views are more popular than they really are.
Another response could be to try to change how we conduct online conversations—rather than having them in one enormous public park, we might silo them in more bounded communities with clearer norms and fixed participants. That may reduce misunderstanding and hijacking, but it comes with risks of its own, like insulation from valid criticism and the development of echo chambers.
I don’t have the answers, and I doubt any philosopher does. But if we refine our philosophical tools in the way I’ve suggested, we will at least have a better sense of the state of play, through a richer understanding of the complexities of our new communicative landscape.
Lucy McDonald
Lucy McDonald is lecturer in philosophy at King’s College London. She works in ethics, philosophy of language, and feminist philosophy. Her chapter “Context Collapse Online” appears in Conversations Online, edited by Patrick J. Connolly, Sanford C. Goldberg, and Jennifer M. Saul (Oxford University Press, 2026).
