Baudrillard and Interiority

Epistemic Status: I really only did a fairly surface level skim of Baudrillard’s actual work mostly cribbed from secondary sources, so I could very well be misrepresenting core things about what he actually said. On the other hand, Baudrillard really just functions here as a jumping off point into my own thinking about postmodernism and interiority, so it might well just be fine.

So I went on another random internet walk and ended up dipping my toes into post-structuralism, especially Baudrillard. For those without arcane knowledge of post-structuralism and Baudrillard, let’s briefly set the stage. Intellectually we start out with structuralism. From the spirit of ‘reading philoosophy backwards’, I’m going to provide an undoubtedly highly simplified and incomplete reading. Before structuralism, people (supposedly) did not think too much about what meaning was, and just assumed that things had inherent meaning¹. Structuralism instead said that meaning is mediated through signs, and that signs themselves do not have inherent meaning, since they are ultimately arbitrary, but rather take meaning from the relationships and differences they have with other signs – i.e. the word ‘cat’ doesn’t intrinsically have to refer to the abstract idea of a cat, it is just some meaningless consonants and vowels. Rather, it is only through its interactions and differences with other signs such as ‘meow, mat, dog’ etc that the sign ‘cat’ derives meaning². Structuralism originated in linguistics where it proved a pretty good theory and served as the grounding for a lot of other work. It also moved into the more pure humanities, where it was generalized to say that not only is meaning in language found between signs, but that the meaning of all kinds of systems can be found only by looking at that system as a whole. For instance, in social structures, parts don’t have meanings in and of themselves but depend on other aspects of the system. To jump ahead slightly, this means things like categories such as ‘sane and insane’, or ‘happy and sad’ or ‘bourgeoisie and proletariat’, or ‘hero and villain’ are not intrinsically meaningful themselves, but rather depend on the existence of the other for their meaning.

We can think of post-structuralism and post-modernism as effectively taking and generalizing this insight further while trying to radicalize and destabilize it. For instance, Derrida focuses primarily on destabilizing meanings within text. For instance, if all signs just point to other signs, there is no fixed meaning, and the entire system, although locally coherent, can in theory drift away from what e.g. the author or others intended and so by close-reading texts you can construct alternative equally valid closed-cycles of meaning from the same substrate. Foucault tried something similar with power relations, claiming that for instance, our conception of ‘madness’ is not some intrinsic thing where some people are mad and some people are not, but is ‘constructed’ both linguistically by our linguistic categories for types of madness such as through schemes like the DSM5 etc, and also civilizationally through actual physical institutions such as hospitals, courts, and sanitoriums. These two elements then provide an interlocking synthesis that ultimately supports the power structure, or rather power itself grows out of these syntheses. You may not think you are mad but you hit these five diagnostic criteria. Where did these criteria come from? Well ultimately some guys with some fancy certificates from some fancy institution created them. If you disagree well, because of the words written in some diagnostic manual, then you will be carted off to some other institution where other guys with fancy pieces of paper with all sorts of words written on them can confine you against your will and exert very real physical power over you. The system of power interlocks linguistic and physical power. As civilization advances power becomes increasingly mediated by such linguistic forms. You are not attacked because some big guy wanted to beat you up and take your stuff. Rather, you are locked away because some set of magical words, spoken and written, allows somebody with power to command the obedience of many others in a system that is utterly vast and far beyond your own scale³.

Reading about post-structuralism and post-modernism in depth has given me much more appreciation that there actually is progress in the humanities. The core ideas of constructedness, Foucauldian genealogy, distinguishing between things-in-themselves and signifiers, thinking about meta-textuality and the position of the author and understanding roles narratively, etc, all of these seem obviously good and correct. Now, the specific inferences that are often drawn from these methods are less good⁴, but this is just a classic case of the fact that negative vision is easy, while positive vision is hard. It is always easier to destroy than create. However, it is clear that the analytical tools and claims of postmodernism definitely constitute intellectual progress^{5 6}.

Anyhow, while Foucault and many post-modernists focus on power, Baudrillard focuses on something more subtle. His primary thinking is around what happens when the system of signs becomes reflexive, not simply reflecting their referents in the world, but start directly shaping it. This is his conception of simulation and simulacra – when signs themselves start becoming the primary object rather than their real referents. Ultimately, we end up in a state of ‘hyperreality, where we produce sign systems that are ‘more real than the real’, and take a larger role in shaping reality than reality itself.

This either sounds mystical or banal, or some combination of both. Today’s internet age is in fact vastly more Baudrillardian than the age when he was writing. A classic example is the Instagram travel photo. Initially the photo was primarily a sign that referred to its referent – i.e. to share the fact that you had been to a particular place and to create a permanent record of that fact for your own memories or to share with others. Over time, however, the photo itself, rather than the experience, becomes the object of status and desire. People start organizing their vacations around producing the perfect photo for Instagram, rather than enjoying the vacation on its own terms and just occasionally pausing to take photos. The Instagram story is the ‘hyperreal’ vacation by which the actual vacation pales in comparison to.

All of this should be highly intuitive to everybody to anybody who hasn’t lived under a rock through the age of social media. This is likely another case of ‘reading philosophy backwards’. The interesting thing though, is the reflexivity point which is not always fully recognized. Not only do signs or images become the symbols of desire, but because of this they ‘reach back’ into reality and start molding reality to fit the forms of the image themselves. The instagrammable holiday destination starts organizing itself to mass-produce its most instagrammable points. Some fashion becomes high status for whatever reason and people start wearing the clothes and affecting the mannerisms of the fashion themselves, thus propagating it, and so on.

Lest we think this happens just in frivolous cultural cases, it also applies extremely broadly to all our forms of coordination. For instance, economic resource allocation is extremely heavily influenced by symbols such as stock prices which can diverge far from the actual ‘real’ situation of a company. In many cases, then the stock prices warp the reality by either granting or denying capital which then impacts the long term real success of the company and the management of the company rationally trends towards optimizing the symbol, the stock price, rather than the reality. Similarly, we can think about education, resumes, personal statements etc here also. These are all ultimately symbolic artefacts. A resume or personal statement begins by treating to stand in as a symbol for the person – what they have done, their strengths, uniqueness, etc. However, once they start being used to make serious decisions, they rapidly become the focus of optimization rather than merely portraying the referent. People’s entire early lives get sucked into the production of the most impressive, glamorous, and symbolically legible resume possible. This continues often throughout people’s entire careers. Their education begins with classes and extracurriculars designed to produce the perfect university application; at university their lives are optimized around the classes, research, and internships needed to produce the perfect resume; in the workforce their career moves are designed to further enhance the resume my padding it with prestigious companies, and so on⁷. This is, of course, extremely similar to Caplan’s education as signalling theory. But it goes a small step further. Rather than people pursuing education just to signal various qualities, instead the nature of the educational system itself, and the actions and incentives of everybody within the system, is warped around producing the perfect signal.

Baudrillard’s insight, as I understand it, is to point out that this increasingly inward-facing symbolic system also becomes causally active. This is naively what you would expect but still is somewhat hard to figure out the consequences of. Specifically, once cultural output and symbols themselves become an optimization target then they recursively affect the next round of cultural/symbolic production. A lot of inward facing output is a response to other inward facing output rather than anything in the external world. Literature builds on other literature; lawyers respond to lawsuits from other lawyers; managers have meetings with other managers, and so forth. This is not all necessarily ultimately traceable to some constraint or feedback from the external reality. Rather,all of this cultural output generates its own ‘eddy currents’. As society becomes more complex, and its interiority grows, these ‘eddy currents’ become the dominating causal mechanism rather than anything from the outside. This is ‘simulation’ and ‘hyperreality’.

So while Baudrillard, and his brand of continental philosophy is often seen as irreducibly mystical, we want to try to make at least certain aspects of his thought seem extremely clear, and even trivial. Firstly, we should not just understand Baudrillard as talking about some idiosyncratic pathologies of his late 20th century age of mass-media. Instead, we should generalize him as effectively communicating the phenomenology of a deeper scaling law of cultural production and the resulting increasing interiority of a civilization as its level of cultural production technology advances. Specifically, by interiority, I mean the degree of interactions, media, civilization and cultural output that are self-referential in some sense. Interiority is when people and systems within a civilization primarily interact with other people and systems within the civilization rather than ‘external reality’ in a direct and unmediated way. The claim is then that as civilization advances technologically, it becomes more complex but the amount of inward-facing complexity increases over outward-facing complexity.

We can think of this from an extremely basic geometric analogy. If we think of civilization as a sphere, the radius of the sphere effectively measures the level of advancements and complexity this civilization has. Then the surface area of the sphere reflects its interactions with the outside world, the non-self-referential aspects of the civilization. The volume however represents the civilization-internal processing, output, energy etc. Obviously, as the sphere grows larger, the ratio of the volume to the surface area must increase.

There are a couple of core mechanisms that cause this. Firstly, simply due to the accumulation of historical time, the total stock of cultural output increases as civilizations live longer. This means that new cultural works must build upon a larger existing corpus and become more inward-facing rather than purely ‘fresh’. Better cultural production and maintenance technology also vastly increases the fraction of total cultural output that is usable to any individual – i.e. printing and then the internet made vast libraries of content viable and ubiquitous which would have been mindboggly expensive previously. Secondly, greater economic surplus driven by technology creates the slack that allows for much greater specialization of roles and the construction of larger internal systems than previously possible. Thirdly, greater communication technologies and increased bandwidth enable a much larger fraction of possible communicative routes through a civilization to be used. As the number of people in a civilization grows, the theoretical number of communication patterns scales combinatorically while the raw number of people scales linearly.

We see this historically. At the beginning everybody was hunter-gatherers interacting with the external world to forage for food as their primary occupation. Obviously there were ‘some’ internal interactions such as ceremonies, rituals, mating, warfare etc but the locus was nearly 100% external. As we reached agriculture there began to be the beginnings of internal apparatuses of coordination and control. Kings, scribes, priests, culture, the beginnings of bureaucracy. More effort and time is focused inwards. Economic specialization develops. Professions such as blacksmiths, potters, weavers, scribes, soldiers, priests etc where the output is to provide tools and support to other people rather than directly earning the means of subsistence from the external world. As technology advances and creates greater surpluses less and less of the population is involved in direct interfacing with the external world and moves into positions of increasingly internally-focused output. We see this in the move from agriculture to industry to services. Obviously these civilizations still need some people interacting with the external world, and in absolute terms greater than ever, to maintain the actual material foundations of civilization, but as a ratio their proportion decreases over time.

This idea is very general across multiple scales. For instance this is true in organizations as well as civilizations as we discussed in the decadence post. In a firm, for instance, in a small startup almost all effort is focused outwards on the external world – building products, talking to customers, generally orienting in the world. As the company becomes bigger it needs internal cooperation and coordination apparatuses – managers, skip-level reporting, CVPs and SVPs and so on. Moreover, it is not just the internal coordination structure, rather even because the company is bigger and has accumulated a bigger stock of IP/code etc, even at the IC level more of the work turns into building on what someone else has done, cooperating with some other team, internal tooling, testing and validating work that has already been completed etc. Final ‘touch’ with the external world must necessarily shrink and this is fundamental and not the fault of any particular management methodology. Similarly ‘corporate politics’ must increase, but what corporate politics means here is simply that more of your interactions and credit assignment depends on internal rather than external factors. You are judged increasingly by what other people within the organization think of you than those on the outside.

Thus, there are two points intrinsic to (my interpretation of) Baudrillard. Firstly, that civilizational scaling, especially of cultural technology, tends to increase the interiority of a culture as it becomes more advanced. In the social and organizational spheres this is people primarily interacting with other people and larger bureaucratic or market systems. In the cultural sphere, this is culture becoming increasingly self-referential, building upon, reacting to, or requiring knowledge of prior culture for intelligibility. The second, and likely more original, Baudrillardian move is to notice that these internal social and cultural systems are not neutral, rather they begin to take on a life and logic of their own, increasingly commandeering causal influence and internal optimization power. This is inevitable. As internal systems begin to allocate resources, credit, rewards etc within the civilization, then obviously optimization power by the local agents within the system will be deployed primarily to affect the outcomes and structures of these systems. Similarly, within culture, the tastes and preferences of cultural consumers are endogenous and are affected by the existing culture, and hence we get a recursive loop of cultural optimization which often behaves unpredictably in practice⁸.

This increasing causal closure makes perfect sense and is rational, as is the increasing reliance upon signs themselves rather than their referents. Because human communication bandwidth is slow and limited (and in fact this is something intrinsic to the combinatorial nature of communication as the number of communicators increases), symbols become highly useful compressed representations for coordination. Because others coordinate around and allocate real resources, credit, rewards etc to those who participate successfully in this ‘symbolic economy’, then it is rational to primarily optimize for symbolic legibility. When everybody does this recursively, then the symbols end up causally mediating and determining real outcomes rather than vice versa – reality ends up being shaped primarily by the symbolic system. I.e. in the hypothetical ‘early’ civilization, reality drives the symbols, later in more interior civilization, reality is primarily driven by symbolic machinations in the interior.

An important point to note is that this increasing interiority is neutral or even positive on-net. It is simply a fact of civilizational scaling. In the decadence post, we explore some of the failure modes that are possible with greater interiority, such as the collapse and perversion of the incentive mechanisms that couple individuals within the civilization to the external world. However, just because there are failure modes does not mean the whole process is a bad thing. Rather, when supported by feedback loops and credit assignment mechanisms correctly coupled to reality, highly interior civilizations are extremely powerful at allocating resources, effort and mental output, coordinated across vast numbers of simultaneously optimizing local agents. This level of coordination is impossible for a less interior civilization to achieve and ultimately is a core component of the positive feedback loop that allows civilizations to develop in the first place, since the coordination enabled by this interiority then produces a greater material surplus which can be used for future interiority.

The broad capitalist economies of advanced countries today almost entirely decouple the vast majority of people from actual encounters with the reality of agricultural and industrial production, and this is not a bad thing. In fact, it is in some sense a fundamental aspect of civilizational advancement to create a surplus large enough that interaction with reality can be minimized for most. Direct contact with reality should certainly not be seen as an intrinsic good. There are dangers to interiority, of course, but in some sense the purpose of technological civilization is to produce enough surplus to pursue higher goals which are not directly constrained by reality. The whole point of values in some sense, is that they have the slack to be arbitrary and not simply reflections of deep constraints of reality, although they must, as always, be shaped by them.

In previous posts we have discussed e.g. the cosmic energies and vast industrial megaprojects that large space-faring civilizations can undertake. The important thing to realize is that while utterly immense, these megaprojects should take a vanishingly small fraction of the total ‘mind-time’ of such civilizations. These civilizations will both be immensely successful at coupling to the external world while possessing utterly mindboggling degrees of interiority and accumulated cultural and intellectual production. The civilizational interior self-goodhearting and other incentive decoupling problems are not necessarily intrinsic to interiority, but are engineering challenges that seem likely to be able to be solved by future civilizations. The computational structures and properties of AGI also make solving these problems much easier than with regular humans, and so this will be a massive boost to the effectiveness of these civilizations⁹. Alignment, in this view, becomes essentially a kind of engineering field of social science – how can we design minds and ultimately systems composed of minds to ensure that they remain correctly coupled to both reality and within the value space that they are intended to serve.

Ultimately, as an engineer, we can begin to view the Baudrillardian representational loop as a kind of feedback control process. Once representations and ‘culture’ start exerting real causal power, which is effectively immediately, we get a closed feedback loop between reality and our representations of reality. This is just what it is, neither good nor bad. Now what is important is how to design and construct systems which correctly couple to the external world. In a world of various local agents inhabiting and comprising these systems, then this basically becomes the alignment and coordination problem at a societal and civilizational level.

The danger, of course, which is one of the decadence mechanisms we discussed earlier, is how easy it is for these feedback loops to end up effectively being self-goodhearted. If we go through the classic OODA loop stages, we can see how our feedback coupling to external reality can become increasingly weak or distorted at every stage.

Observe: As we increase interiority, what we are primarily observing is other aspects of the interior ecosystem – signs created by others. Moreover, even if we observe reality directly, then reality is also increasingly shaped by the causal actions of the internal mechanisms of our civilization also. Observation, rationally, becomes more social. What are other people thinking or doing? Should I follow them or be contrarian? What do others say about this event? What does the news or social media say? There are obvious benefits to this observation system. Media produced by others is often vastly more information-dense and generally contains superior synthesis to raw data from reality. A single person cannot hope to process, understand, and evaluate all of the raw data that ultimately produced the media they consume. Rather, the synthesis and compression pipeline of media and broader culture is vital to understanding anything at the speed and breadth which is required to function effectively in modern society. However, clearly, there are trade-offs. Every act of synthesis or compression can distort, misinterpret, or mislead either deliberately or accidentally. Raw data is extremely high variance but low(er) bias. Cultural media leans the opposite way towards lower variance but higher bias. This can become distorting over time. Moreover, due to the reflexive nature of the self-referential loop of culture, the bias can compound itself. Sometimes this is through deliberate group-think or even targeted influence campaigns. More generally it is just a natural consequence of biases becoming self-entrenching. Once there is a bias, others internalize and produce other outputs which assume the same bias and perhaps even deepen it, and then this produces a whole packet of works which all look like independent confirmations of the same direction, but in fact all share a hidden mediating bias.

Orient: We increasingly orient not necessarily to direct observations but to the signs, concepts, and other cultural latents produced by our own civilization. This is natural since we would hope that all of this cultural and intellectual production produces good frames and compressions and syntheses and models for understanding the world. However, it provides correlated failure modes. If everybody is understanding the world using the same concepts and ontology, then this means there are important things that can be missed where these ontologies fail or mislead. Moreover, increasing numbers of concepts and judgements refer to social or internal effects rather than the raw percepts themselves. The trade-offs here are similar. Nobody can meaningfully hope to grasp the complexity of the real world through pure individual a-priori reasoning. Using frames, ontologies, and schemas built by others to orient is necessary to function in a complex cultured civilization. However, these shared ontologies obviously create correlated biases which can cause cascading failures in the cases where the ontologies fail.

Decide: There are two mechanisms here. Firstly, decisions are increasingly driven by social and cultural pressures such as expectations, judgement, impact on others’ etc. Decisions are also obviously highly affected by the conceptual schemas and ontologies used to understand the world and action space, which are increasingly inward-facing. Secondly, decision-making becomes increasingly reflexive – in that the primary target and output of the decision is not necessarily the actual external world but rather other elements of the social/cultural/symbolic system as a whole. In fact, many decisions end up almost entirely focused inwards. This means that the success or failure of decisions often comes not from modelling a fixed external system with known laws but rather the recursive equilibrium of many different agents interacting. Economics gives many good and famous examples of this where speculative market prices often end up far less about the actual long-run cash-flows of some particular asset and vastly more about recursively modelling what other agents are likely to believe that other agents are going to pay for it in the future. Importantly, understanding such reflexivity is not optional for effective action. In fact, it is often the most important component to understanding whether a given action is likely to have the desired effects or not. However, this kind of reflexive decision-making can produce extremely strong large-scale correlations based on effectively no true information. Economics again provides plenty of examples of mass speculative euphorias followed by mass panics and collapses. These are part of the intrinsic dynamics of reflexive decision-making, since they are functions of positive feedback loops which are locally rational at every step. Ultimately, this kind of reflexive logic will come to play a substantial role in highly interior civilizations and often results in decision-making that from the outside is highly unpredictable and unusual, but which is relatively rational from the inside.

Act: The action space itself is also increasingly constrained by, or focused on, interior-facing actions – e.g. actions primarily involve coordinating with other individuals and groups, achieving consensus within structured group decision-making settings, or producing cultural outputs designed to be legible to others within the same cultural context to influence them in some way. These kinds of reflexively-oriented actions are necessary to be able to pursue any kind of highly coordinated action or strategy across a large group of agents. However, at the same time, often coordination between agents can be applied directly against the machinery of reward and credit feedback loops that keep the civilization correctly coupled to reality – for instance in primarily seeking to manage perception of others, and in all the various types of possible collusion between subsets of agents. From the collusion standpoint, a core challenge of mechanism design is to figure out how to enable certain kinds of coordination while disfavouring others as much as possible.

These reflexive dynamics can produce interesting behaviours, cycles, and equilibria. One interesting one is the self-reinforcing Schelling point. This is where some particular element of the system, usually due to some arbitrary initial or historical condition, becomes highly important as a Schelling point which then becomes self-sustaining. This Schelling point is a property entirely of the symbolic system itself, and is usually due to a need for legible coordination among many entities. A classic example of this is prestige. For instance, consider prestigious universities? Why are they prestigious? Well they have the best students and produce the best research? Why do they have the best students and research? Because the best people apply to study or work there? Why do the best people apply to work there? Because it is prestigious. The loop closes. Moreover, it is usually self-stabilizing. The prestigious institution is not just an ephemeral point of coordination, it can use this point of coordination to bootstrap a very real material infrastructure which it can use to entrench itself even deeper as a Schelling point¹⁰.

We can also get limit-cycle dynamics such as oscillations. Fashion is a classic case of this where whatever was trendy last season is terminally uncool this season, while old things get periodically recycled by being ‘retro’. Academia also goes through similar oscillations but on a much slower generational timescale. More generally, in many different cultural fields there is a pressure to be ‘original’ in a sense of reacting against and charting entirely novel regions of the ‘latent space’ than your immediate predecessors¹¹. If the latent space is constrained to only a few attractors, we end up with a punctuated equilibrium of jumps between these attractors. If the latent space is large and extremely varied, like we see in e.g. art and music, then we tend to see a kind of branching random-looking exploration happening in bursts. This is interestingly different to the cumulative parts of science academia, where originality is also demanded, but it can and indeed must be a deepening sort of originality. In physics, the idea isn’t to deny what Newton and Einstein said and invent some totally new crazy physics. Rather, it is to continue a cumulative process of thoroughly exploring ever deeper branches of the same shared tree.

This OODA loop breakdown then provides us basically a schematic of various kinds of pathologies that can occur in these kinds of self-reflexive systems. These include perceptual goodhearting whereby you only see what you want to see, which prevents taking meaningful action, and more broadly the perceptual environment and ontologies becoming highly correlated since even if your perception remains unbiased, you end up surrounded by, and primarily observing, others who are affected by the same biases that you are, so it becomes self-reinforcing. This applies both to the raw percepts themselves and also, more subtly, to the ontologies and concepts used to understand the raw percepts, since the ontologies are usually transmitted culturally themselves vs derived from pure independent reasoning. Similarly, when deciding and acting, we get the core reward-hacking/goodhearting failure modes of a.) trying to directly act to modify the reward and credit assignment machinery, b.) trying to act primarily to influence others’ perception and hence affect their behaviour which ultimately feeds back into your perceptual system. A classic example of this is organizations trying to manage their critics directly rather than the actual underlying factors which led to the criticisms in the first place. This is often highly tempting because attempting to manage perceptions is often much easier than managing the actual underlying reality.

Overall, then, being able to catalogue and come up with a proper theory of all the kinds of interesting reflexive dynamics we can get inside highly interior civilizations is thus a deep and very open-ended research program. While Baudrillard, in his own way, may have begun to open the door, he has hardly begun to map the giant palace of interesting phenomena he stumbled upon. Obviously different fields such as nonlinear dynamics, economics, cultural studies, etc have their own views of these dynamics and the challenge will be to figure out a unified framework to place all of these individual phenomena and then figure out, to the extent that this is possible, the mathematical laws governing them and then the control affordances that these laws provide at different scales. This is certainly an interesting direction to go in.