Power and Common Knowledge
October 27, 2021
In this post I want to unravel some thoughts I’ve been mulling over for some time, concerning the relationship between critical approaches to power and the decision theoretic concept of ‘common knowledge.’ I’ve touched on this before in relation to advertising, and was delighted to hear it brought up by Evan McMullen on the Meta-Ideological podcast. This relationship seems to me to provide the key to understanding the deep sense of political and cultural inertia that is so characteristic of the present moment. One approach to this inertia is to consider it as a deterioration of collective agency, or a kind of generalised coordination failure. This deterioration can in turn be made precise by analysing it as a failure of common knowledge production. In this way, deep questions about political horizons can be linked to the kind of media theoretic concerns that impinge on common knowledge. (I won’t rehash any of this here, but a detailed account of the relationship between common knowledge and coordination can be found in this LessWrong post.)
In the discussion, the point made was that power is often exercised through ‘information blocking’ tactics. The given example was workplace taboos, where saying publicly the obvious thing that everyone knows privately could get you fired. This barrier stops the tabooed fact from becoming common knowledge even though it is mutual knowledge (i.e. everyone knows it, but they don’t know that each other knows it, or know that they each know that they each know it, etc), and this is sufficient to undermine their capacity to coordinate their actions. The situation therefore remains unchanged, despite an overwhelming shared interest in changing it. The implicit suggestion seemed to be that a politically empowered society would be one in which information blocking is minimised, because this would be a society in which the capacity to generate common knowledge is maximised. (And here ‘society’ could be swapped out for smaller units of organisation if preferred.)
The point I want to make is that while information blocking does represent one form of strategic power, it is by no means the only one. This is because there are other ways of inhibiting common knowledge production. Moreover, I would argue that information blocking is not the dominant form of power in operation today. Once the alternatives have been appreciated, I believe this should leave us wary of claims that straightforwardly link increased collective agency to the liberation of information flows. The comments in the discussion were brief, so I do not know whether the line of thought I’ll pursue here is in tension with or perpendicular to Evan McMullen’s position—at any rate, there is a whole cluster of issues here that are well worth exploring.
1. Power Trips
The point can be approached by first considering the notions of ‘sovereign’ and ‘disciplinary’ power, theorised famously by Foucault (Foucault, 1995). For Foucault, these categories represented two fundamentally different forms of power—sovereign power is an exercise in sheer coercive force, while a disciplinary apparatus operates by strategically shaping the norms that individuals apply to themselves. But in his excellent book Rational Ritual, Michael Chwe uses the concept of common knowledge to show that these two forms of power are not as different as they appear to be (Chwe, 2001).
First, sovereign power is never a simple case of do-this-or-else: even where a sovereign engages in brutal forms of violent suppression their ability to do so depends on commanding a legitimate authority among those who underwrite the sovereign’s force—typically a military or an aristocracy. This is a coordination problem, and depends on the sovereign establishing and maintaining their legitimacy as a point of common knowledge amongst (at least) an inner circle. Traditionally this has been achieved by monopolising public space: huge monuments, lavish processions, etc—the historical ubiquity of which would be difficult to explain on an account that reduces sovereign power to pure coercion. And of course this authority can fail, as Charles I discovered.
Second, in the case of disciplinary power we can note that the efficacy of a norm—insofar as the efficacy of a norm is measured by its capacity to coordinate action across a society—is not a function of an individual’s direct application of it to themselves, but rather of their perception that other people are conforming to it. (This is similar to the point that dumbed-down political messaging often succeeds not because people are incapable of grasping nuance, but because it easier to be confident that other people have grasped a message when it is dumbed down, and to be confident that they have grasped that you have grasped it, etc—this second factor is what is required for coordination). When we say that the disciplinary apparatus shapes norms, then, we do not need to propose that it causally effects a change in individuals’ immediate self-relation (not even in a highly decentralised, ‘micro-physical’ manner, as Foucault had it), but rather that it operates by distorting the medium of practical reason, in effect shaping the space of possible coordinations by controlling the information each individual has about what others take the social rules to be.
On this interpretation, both sovereign and disciplinary power depend on a strategic deployment of common knowledge—the fact that the first achieves this through a monopoly on public broadcast while the latter achieves it through a homogenisation of personal experience does not change the fact that they are based on the same underlying mechanism. Both install items of common ‘knowledge’ that manipulate coordination possibilities in their favour, and thereby define a communicative regime of taboos and obligatory performances. Both forms of power therefore fall under the pattern named by information blocking. The important lesson to draw from these reflections is that information blocking does not function by inhibiting the production of common knowledge per se, but rather by stabilising common knowledge in some configuration which works in the interests of the incumbent regime, even when mutual knowledge (what individuals actually believe) diverges from it. Taboos are ossified solutions to coordination problems—though often to the wrong ones, and often asymmetrically privileging particular interests.
Chwe’s argument—that power should be understood in terms of the rational manipulation of common knowledge rather than as a direct (i.e. a-rational) causal influence—can be seen as a decision theoretic analogue of Baudrillard’s earlier critique of Foucault. For Baudrillard, the problem with Foucault’s conception of power is that “it substitutes a negative, reactive, and transcendental conception of power which is founded on interdiction and law for a positive, active, and immanent conception.” (Baudrillard, 2007, pp. 35–36) In this immanent conception power is treated as a kind of pulsion, a force which in effect produces the social, initially concentrated in the figure of the sovereign before being micro-physically distributed across the social body in its disciplinary configuration. According to this picture norms, laws, and interdictions are essentially epiphenomenal in relation to power—transcendent alibis for its true operation, which is immanent and causal. In contrast, Baudrillard’s position is that power does not, strictly speaking, exist. Or more accurately: power is an imaginary effect, something socially exchanged, which is to say it exists only insofar as it is given and received. On this account norms, laws and interdictions are not epiphenomenal—rather they are authoritative in practice precisely to the extent that they are taken to be authoritative. Power is in fact always reversible, but it will function irreversibly to the extent that it is collectively imagined as irreversible. In this sense it is Foucault’s conception of power as a kind of cybernetic causality which provides an alibi for the true operation of power, which consists rather in the strategic manipulation of its imaginary (i.e. of common knowledge).
The above reflections gesture to another, wholly different form that power can take: the pure destruction of common knowledge, or the outright inhibition of the production of any common knowledge at all. The result of this would be a radically uncoordinated population, a collection of independent atoms lost in Brownian motion. Such a population, because uncoordinated, would be infinitely susceptible to various kinds of statistical nudging and modulation—techniques collected by Deleuze under the heading of ‘control societies,’ and explicitly identified as a new regime of power distinct from and in the process of replacing Foucault’s disciplinary apparatuses (Deleuze, 1992). The destruction of common knowledge per se is not achieved by information blocking, but by information flooding—this, I would suggest, is the primary source of the political disempowerment endemic today. This is also why we should be wary of attempts to naively liberate or unblock information flows—they risk playing directly into the hands a regime of power for whom information transparency, rapid dissemination, and ultimately a paralysing excess, is the operative principle.
2. The Divine Plight of Kings
The idea is fundamentally very simple: the more information there is floating around in public space the harder it becomes to know what other people know (even if it’s all good information), i.e. the harder it will be to establish common knowledge. There’s nothing mysterious about this: it’s just combinatorics. But before unpacking what this means in practice, it’s worth clarifying a few things. Firstly, I’ve started putting scare quotes around the ‘knowledge’ in common ‘knowledge,’ because to play the role it does in coordination, common ‘knowledge’ does not even have to be true. On this basis it might be tempting to start speaking about ‘common belief’ instead, but as it turns out—and this is even weirder—to play the role it does in coordination, common ‘knowledge’ does not even have to be believed. The point is made succinctly by David Lewis:
If yesterday I told you a story about people who got separated in the subway and happened to meet again at Charles Street, and today we get separated in the same way, we might independently decide to go and wait at Charles Street. It makes no difference whether the story I told you was true, or whether you thought it was, or whether I thought it was, or even whether I claimed it was. A fictive precedent would be as effective as an actual one. (Lewis, 1969)
Insofar as the ‘social imaginary’ can be taken to denote that set of narrative tropes, practices and shibboleths which defines a culture and its deepest coordination mechanisms, nothing in its ability to do so depends necessarily on anyone considering its contents to be true—a fiction is no less socially efficacious when recognised as a fiction, just so long as it is recognised as a shared fiction (it’s worth comparing this with Lyotard’s account of the self-legitimating character of narrative knowledge in The Postmodern Condition). This helps to understand how it is possible for common knowledge and mutual knowledge to diverge, sometimes radically, even while common knowledge remains the locus of de facto power.
As an example, let’s consider the divine right of kings. Let’s imagine that the king is obviously insane and everyone knows it, and furthermore that no-one really believes in the divine right of kings anymore. Nevertheless, the king will still retain de facto authority (which is to say, people will still be responsive to his commands) so long as no-one points out publicly what everyone knows privately—until they do, the divine right of kings remains an equilibrium point of common knowledge, even though no-one believes it and even though it’s becoming a nuisance. But there’s a good chance that no-one will do this, at least not unless things get really dire. To say the obvious publicly is deeply risky until you can be sure that other people not only believe it, but are prepared to say so and act on it, and that they believe that other people believe it, and so on—but you can’t possibly know all this until someone says it publicly. The structure of the dilemma is similar to a Mexican stand-off: everyone wants everyone to put their guns down, but no-one wants to be the first to do so, so no-one does.
Power here is maintained by information blocking: the taboo on questioning the authority of the king serves to preserve its divine right as an item of common knowledge, and therefore its status as the effective locus of coordination. This is true even though mutual knowledge has drifted considerably, and even directly contradicts common knowledge. Now let’s reverse this situation and consider things from the point of view of the king. Let’s imagine the king never once wanted to be king, and has long resented the fact of his birth. In fact the king’s ‘insanity’ and erratic behaviour is no more than an expression of the contempt he feels for this social cage in which he has found himself, which offers no possible exit or pathway to abdication.
But wait—is this right? Is it true that there’s no exit? The king is after all the sovereign, the one person whose authority is in theory not conditional on any corresponding responsibility—among other things, the king is not bound by the kingdom’s taboos. So why can’t the king just be the one who points out that his divine authority is a fiction? Wouldn’t this be the public act that brings common knowledge into line with mutual knowledge, thus dissolving the power structure? If the king’s power truly consists in the exclusion of this information from public airwaves, then surely it does not matter who broadcasts it?
Well, no. If the king stands up and says “the divine right of kings is a load of shit actually” at the joust, this will most likely fail to establish it as common knowledge. Why? Because the message has bad metadata. If one of the king’s subjects had said it, then they would be doing something inherently risky, as has been mentioned—but it is precisely this risk, perceived by others, which would establish the message as credible, which is to say as a reliable indicator of motive: specifically the motive to speak sincerely and be prepared to act on it. There is no other reason to take such a risk. But if the king says it then he is taking no such risk, and indeed there are many plausible motivations for this utterance—perhaps he is testing his subjects’ loyalty, or playing some weird psychopathic power game. Given this uncertainty, the safe option for the subjects is just to ignore it or shrug it off. Haha good one Sire. Everyone knows that everyone else perceives this uncertainty, and so the risk of agreeing with the king remains—the subjects are still very much in a Mexican stand-off. From the perspective of the king, who means what he says, this all comes across as a strange form of gaslighting through flattery. Ultimately, he too remains powerless to change the situation.
The conviction motivating this post (Baudrillardian through and through) is that the disempowerment distinctive of contemporary inertia has far more in common with the disempowerment of the king than with the disempowerment of his subjects. We do not live in a world where power is wielded under the sign of omerta. It simply does not need to be–it is far more efficient than that. We live, rather, in a world where the ‘taboos’ of power are constantly and incessantly being pointed out, and in response everyone goes “haha good one”—then carries on as before.
3. Bad Metadata
What does it mean for a message to have ‘bad metadata’? What even is metadata in this context? Consider the difference between seeing an ad during the Super Bowl and encountering the same ad randomly on the internet. Their content is identical, but their metadata is different: when you see it during the Super Bowl you also know that many other people are seeing it, whereas on the internet you have no idea. Watching a Super Bowl ad is like receiving an email that you can see has been cc’d to 90 million other people. Metadata is independent of content, containing only information about context. This 2nd-order, contextual information often has critical implications for common knowledge and coordination. Chwe, for example, argues that the reason advertising spots at the Super Bowl are so sought after is precisely because of this metadata: knowing that many others have received the same information as you is crucial if you are considering buying, say, a new communications technology, something which will be useless unless many others also buy it (here we might think of Apple’s famous Super Bowl commercial from 1984).
Making the distinction between content and metadata highlights how thinking about power in terms of information blocks and flows can be misleading: it tempts us to place focus on content, when often it is the metadata of a message that occupies the critical position in relation to common knowledge and coordination. This tension is played out around us everyday. The same logic that identifies information blocking as the root of political power is the same that sees political and media strategy as primarily a matter of ‘signal boosting’—both in terms of raising awareness of one’s own cause, and in construing any action which implicitly grants a platform to an opponent as complicit in their strategy. According to this logic, it can only be a good thing when McDonald’s releases its Black Lives Matter commercial—yes it may be cynical, an act wholly conditioned by and subordinate to the profit principle, but surely what is important is raising the visibility of the message in public space (because isn’t this, and this alone, what creates common knowledge?). And as far as sheer reach goes, you couldn’t do better than McDonald’s. If the suppression of anti-racist struggle truly consists in the exclusion of its message from public airwaves, then surely it does not matter who broadcasts it?
Well, no: this message clearly has bad metadata. A successful public intervention would be one that produces common knowledge of a shared political will—a belief in an outcome and a capacity to act on it. But this depends critically on the credibility of the intervention, a credibility which in this case is undermined by the ambiguity of motive. McDonald’s public intervention does not signal an unconditional commitment to the cause, it signals a highly conditional judgement that signalling allegiance to this cause is good for business right now. Its metadata is corrupted: if you start echoing and sharing (i.e. signal boosting) it then you only corrupt your own metadata, becoming, to use a phrase of Lyotard’s, “a programmed effect in the opponent’s strategy.” (Lyotard, 1986, p. 16) In this dynamic we can see what is obscured by the logic of signal boosting: that when a boost in signal is accompanied by a corruption of metadata, this actively undermines its capacity to produce common knowledge, and thus to facilitate political coordination.
And this, I would submit, is the whole story. The contemporary regime of disempowerment is governed by the following equation: the inverse relationship between the strength of the signal and the integrity of its metadata. This trade-off is implemented in practice by an incentive structure: cash prizes for unmasking the taboos of power, for finding and airing the skeletons in the closet. The cash prize is, of course, the very thing that corrupts the metadata of the act of unmasking. The result is both a flood of information—and we can assume all this information is true, nothing in this argument hinges on disinformation—and a universal scrambling of metadata. The final effect is something like a sustained sybil attack on the social imaginary, fundamentally undermining the production of common knowledge through sheer combinatoric explosion. We might think of that Fredric Jameson quote, more often associated with Mark Fisher: “it is easier to imagine the end of the world than to imagine the end of capitalism.” Taken at face value, this is of course nonsense—a quick scan around Netflix or DeviantArt reveals an endless conveyor belt of post-capitalist utopias and dystopias being imagined every day. But it taps into a truth in the sense that what seems impossible now is to coordinate around an imagined future, to take any one of them seriously enough to enact. If we look closely at this, I think we will discover it is a problem of excess rather than lack: imaginings of the future bifurcate and split into new variations so fast that building the common knowledge required to collectively commit to any one of them has become combinatorially intractable.
In this environment, to base a strategy purely on signal boosting or information unblocking is to fall into a trap. To boost your signal, you have to conform to the codes or protocols of public space—a concession to the incentive structure which in practice means pursuing those cash prizes (or appearing to pursue them, which amounts to same thing as far as metadata is concerned—more on bad incentives here). Rather, any political movement must place the preservation of metadata integrity at the centre of its media strategy. This will mean becoming invisible.
References
- Baudrillard, J. (2007). Forget Foucault (P. Beitchman, Tran.). Semiotext(e). [PDF]
- Chwe, M. (2001). Rational Ritual: Culture, Coordination, and Common Knowledge. Princeton University Press.
- Deleuze, G. (1992). Postscript on the Societies of Control. October, 59, 3–7. [PDF]
- Foucault, M. (1995). Discipline and Punish: The Birth of Prison (A. Sheridan, Tran.). Vintage. [PDF]
- Lewis, D. (1969). Convention: A Philosophical Study. Cambridge. [PDF]
- Lyotard, J.-F. (1986). The Postmodern Condition: A Report on Knowledge. Manchester University Press.