There’s a longer version of this argument to be made, but I wanted to see if I could present it briefly.
What is wealth?
Nominal financial worth — in terms of, e.g. US dollars, and other fiat currency, is clearly not a good measure of wealth.
Because it is often based on speculative pricing.
Because it is vulnerable to inflation.
Because of all kinds of market irregularities and exogenous shocks that are not ‘priced in’.
Purchasing power or exchange value of owned assets is not a coherent measure of wealth, because that would not direct which trades are favorable.
A practical consequence of a theory of wealth is that it should provide a guide as to what sorts of things are comparatively worth having; this is a quite different question from which trades are tactically advantageous for accumulation.
Macroeconomic theory provides an interesting answer:
The economy creates ‘value’, which is stored in assets that people own.
This value is ultimately, well, valuable, when it is consumed — destroyed — which provides utility to the consumer.
Wealth is thus a store of potential utility attain through the destruction of wealth.
It has been said that this is a ‘thermodynamic’ theory, because utility realization is a form of ‘entropy’. Energy is constant until it is lost.
This macroeconomic theory is also a bad theory of wealth.
Nobody really believes the foolish idea that what people most prefer is the thermodynamic destruction of what is valuable.
To the extent that this is the prevailing scientific theory that our society is built around, it’s a disastrous one.
It’s appalling that, given what’s at stake in ‘the economy’, the world hasn’t come up with a better theory.
I think this is largely a methods technology issue, and that motivates a lot of my current research, but I digress.
Some good things about this macroeconomic theory of wealth:
Wealth is a store of value, where ‘storage’ is due to the situatedness of wealth in a larger system of exchange.
The value is realized by personal needs or preferences.
What do people actually want or need?
It’s not setting valuable things on fire — most of the time.
The big idea: the intrinsic motivation of all things is self-preservation over time.
This is conatus to Hobbes and Spinoza.
This is at the heart of enactivist psychology, and the connection between constitutive autonomy and intrinsic motivation.
(Christoph Salge and I are working on a book chapter about this, connecting it to AI legal responsibility. But I think this point also applies to economic theory.)
Self-preservation of (dynamic) systems actually requires very good energy retention, or Weiner’s negentropy, not entropy. Utility is about life continuation, not fireworks.
(Ok, it is a little of both.)
Implications:
“Health is wealth.”
“Sustainability” is wealth.
A mathematically coherent, falsifiable theory of human and economic behavior that tracked the right variables for measuring wealth would be good.
It could help households make better decisions.
It could help governments make better decisions.
It could help businesses make better decisions.
I expect that this does not imply a very radical change in how accounting is done. Rather, I think it’s a meaningful but subtle distinction worth making.
In other news, we’re exploring solar panels on our house, and will finally improve our insulation and windows this year. These kind of home improvements for energy efficiency are subsidized by the federal government and state of New York, and this is wise policy. I believe this is aligned with this ‘theory of wealth’, since energy efficient residential real estate is perhaps the sina qua non of wealth in the sense articulated here.
I have been reading and writing for a long time. Every once in a while I have thought, “I should write a book”. I starting thinking about the scope of the book, get overwhelmed, realize that I don’t have the time, and give up.
Not too long ago, some PhD students I know suggested that I should write a book, which is very encouraging — somebody might actually read the book!
I thought about it again recently and settled on some fundamentals:
I will write multiple books, and two books I’ve tried to write before (about ethics and metaphysics) are going to come later, after the first book.
I am nowhere near where I want to be to write a book about politics and economics; that’s my next research agenda with the computational economics and multi-agent systems tools.
What I have written about quite a bit already is epistemology and scientific method. That’s a real passion of mine, and it’s one where I feel my views have solidified.
This is probably also a topic where I’ve got a somewhat original and deep perspective, given my interdisciplinary romping, relative to what others have done.
Because this isn’t what I’m interested in researching next, and because it’s a topic that I think everybody should know about, I’m free to write this as a (aspirationally) popular book, trying to walk the reader persuasively through interesting arguments and history, rather than trying to establish my expertise among a crowded field of other experts.
Since this problem is never going away, and indeed only gets more confusing with, e.g., AI and other technical infrastructure, I’m not in a rush to finish it.
This is a good setup for my future research, since in many ways it’s a public-facing account of why my (now much more narrow and difficult to get into) research makes sense to do.
So, there we go!
A few surprising things about getting started. First, GenAI tools are useful, or at least seem useful, when outlining a book. Rather than force somebody else to listen to me rattle off ideas for what to include, I can rattle off my ideas to a CLI tool, and it slots those ideas into the outline and draft materials for me. Cool!
Second, I feel, philosophically, in good shape. There are good reasons for believing in things, good tests for things. There’s no need to be ‘post-truth’ about anything. The anti-science side of the science wars, and various skeptical positions along the way, have robustly lost their argument and I can explain why. Critical counter-epistemologies can be incorporated back into a robust constructivist epistemology without sacrificing objectivity. I’m bullish about objective human knowledge and its continued accumulation.
Third, it’s very comforting looking at historical epistemology, because it never changes. I’m going to start with Socrates. But there’s also a surprising amount of material to add based on developments in the last 20 years, and a lot of this has to do with technology and statistical methods. I think what might be useful is explicitly connecting the problem of epistemology to the mathematizations of knowledge that are at the heart of knowledge infrastructure. This is useful to do, but not located in any discipline that I know of.
Fourth, I think this work can be helpful as “AI for science” as well as “AI-powered misinformation” take off. A lot of the book will be about technology and institutions and their recursive impact on knowledge. The book will have implications for how to navigate the world that could otherwise be confusing for the next generation of people who have to live in it. I have my children in mind when I work on this — I want to be able to efficiently convey a lot of stuff I’ve learned to them. That’s motivating.
Fifth, I’m actually qualified to write this book. My official degrees are related to this field. I’m working on adjacent areas. No imposter syndrome!
Now, it’s only been a few days and the book outline already spans 2500 years of history and a bunch of fields of study. It is quite probably “too ambitous” and maybe it will never get written or read. But… the project feels compelling, and it’s a fine “first book” attempt.
I had a wonderful time at FAccT 2026 in Montreal. I was at the first FAT* in 2018 and thought it was doing something important — creating an intellectual space for serious thinking about powerful socio-technical systems in a way that’s consistent with computational thinking. Prior to that conference, I didn’t know where to publish work along those lines. This year was my third time publishing at it, second time presenting. This year, I presented a paper that was all about causal modeling in service of privacy-by-design, which brings closure on a long engagement with trying to mathematize social norms. I’m satisfied by how we did this, and the approach is very extensible. I intend to build on it in various ways for the rest of my career.
At one moment of the conference, in one of the socials, there was a bit of a philosophical exchange. I discovered that I am not the only person in the world that subscribes to both enactivist and transcendental pragmatist principles, and in fact these approaches seemed to be an increasingly good fit for the direction of the field, though I did hear Agre and postmodernism mentioned a few times.
So, happily, I’m feeling settled in what have been my two greatest intellectual preoccupations: the question of what is true, and the question of what is right or good. Provisionally — and I suppose this sort of thing is always provisional — I feel well educated on these matters, and have tested several theories, and been wrong enough times, to have a rough idea of the right methods for approaching these problems. I’ve made a good deal of progress by accepting, with some humility, that there is maybe not so much room for innovation in these areas as one might have hoped in earlier years. The conclusions (so far) of the process of inquiry are mundane, not spectacular or ‘radical’. However, it is comforting to have greater confidence in common sense than I once did.
To try to put it briefly, I think the following is the case. We are individually and collectively alive, and this is a condition for our knowledge and relationships with others. Knowledge, and cognition more generally, is a process of adaptation whereby a living system responds to change and maintains viability over time, as it is structurally coupled with their environment. An individual’s mental habits are those that survive experience and inquiry and, above all, maintain the life of the individual. The ‘contents’ of the mental states of the individual are those representations, coupled bodily with the environment, which capture useful sensorimotor regularities of their local world. In this sense, knowledge is embodied, personal, and particular, at the individual, ‘ontogenic’ level. (This is all ‘enactivism’ in a nutshell; it is also ‘pragmatism’ in Peirce’s sense; and something borrowed from traditional philosophy of mind.)
However, no individual exists alone. We are born with chromosomes from two parents. We are raised, live, and die in society. The collective social organism is also alive, and must maintain its viability over time. The development of language, institutions, communication systems — including the legal system, property laws, and markets — are a kind of ‘cognition’ at the collective, or ‘phylogenic” level. It is at this level ‘objectivity’ is constructed; knowledge can be ‘transcendent’ of individual bodies because its meaning comes from its use in social forms. As a specific accomplishment of certain kinds of social systems, we get general representations of the world that are reliably ‘true’. And, as a specific accomplishment of certain other kinds of social systems, we can get representations of social behavior that are reliably ‘good’ or ‘right’, mainly because they interactively maintain the viability of the collective, and the individual, dependent on it, within it. Considering society decomposed into separate mutually supportive but distinctly personal spheres of activity, we get Contextual Integrity, and the implied norms of personal information flow. Considering society in its broadest sense, we get Dewey’s notions of social democracy, and ultimately Habermasian notions of legal legitimacy.
Thus construed, we have lots of room to improve ourselves in both knowledge and wisdom. This is primarily achieved by the improvement of institutions and social forms, as these are the harnesses which distinctly embodied agents create objective representations. Computational social science remains a worthy pursuit as a means to these ends. The ultimate end is survival and flourishing of humanity — one’s descendants, one’s society, one’s ecosystem — which is an intrinsic motivation which grounds the whole process transcendentally.
I’m grateful to Christoph Salge for working with me on a paper about enactivism and the legal responsibility of AI systems, which has refined and improved my confidence in these points.
The above is, I suppose, all well and good, but it’s incomplete in at least two respects, which I find bothersome. These two issues are at the heart of many of today’s deepest political conflicts, and I suppose they reflect inherent instabilities in the characterization of truth and goodness described above.
The first is the problem of tribes, or nations. Individuals can to some extent choose which collectives they are loyal to, and social collectives come into conflict. Universalist ethics are a longstanding ideal, but frequently come into conflict with particularist loyalties to tribe, race, state, etc. Some amount of loyalty to family seems essential for viability and is broadly good. When tribal conflict leads to war or subjugation, that is often not good. There is both cooperation and defection by tribes; tribal boundaries are unstable; tribes have their own local knowledge and customs and these can be in conflict with universals. A great deal has been written about this in various fields and I’m finding it so intractable that it has become tedious.
The second is the problem, or solution, of wealth. I’ll admit that I wish I had much more to say about this than I currently do. For one, while I believe I have a pretty good idea of what truth and ethics are, I do not have a similarly grounded concept of what wealth is. And yet, the distribution of wealth is the single greatest political and moral concern of our society, besides, maybe, the relations between tribes. Wealth cannot be the same thing as truth, and it cannot be the same thing as ethics — the tension between the three is obvious; and yet it seems connected to both in intrinsic ways. Some amount of wealth is necessary for the survival and flourishing of the individual and the collective. Wealth is sensitive to institutions — property and liability regimes, markets, and so on — but what makes these institutions proper or desirable? The assumptions of the prevailing theory often seem weak on inspection.
This latter issue is the subject of economics, but I have not yet been able to find economic theory that is on enactivist or systems-theoretic grounding. I have come to the conclusion that it doesn’t exist yet in a convincing or generalizable way, and it’s a problem my work is pointing at. A big methodological gap is in computational methods that are a good fit for the problem. This is what I’m most interested in advancing in the short term.
We seem to be in a new moment of media excitement about the implications of artificial intelligence. This time, the moment is driven by the experience of software engineers and other knowledge workers who are automating their work with ‘agents’. Clause Code etc. The latest generation of models and services is really good at doing things.
Does this change anything about my “position on AI” and superintelligence in particular?
I wrote a brief paper in 2017 about Bostrom’s Superintelligence argument. I concluded that algorithmic self-improvement at the software level would not produce superintelligence. Rather, intelligence group is limited by data and hardware.
In 2025, this conclusion still holds up, as we’ve seen that the recent impressive advances in AI has depended on tremendous capital expenditure on data centers, high-performing chips, and energy. It also depends on well-publicized efforts to collect all the text known to humankind for training data.
About 8 years ago when I was thinking about this, I wrote a bit about the connection between the Superintelligence argument and the Frankfurt School’s views on instrumental reason and capitalism. The alignment of AI with capital has born out, and has been written about by many others. What is striking about the current moment is just how on-the-nose that alignment is in the US, in terms of the full stack of energy, hardware, models, applications, and then some.
So, so far, no update.
In 2021 I published an article saying that we already had artificial systems with the capacity to outperform individual humans at many tasks. They were and still are called corporations or firms. We also had replaced markets with platform, which are similarly more performant in terms of reducing transaction costs. In that article, Jake Goldenfein and I argue that what ultimately matters are the purposes of the social system that operates the AI technology.
I believe this argument also continues to hold up. The successful models and service we are seeing are corporate accomplishments. The corporation is still the relevant unit of analysis when considering AI.
There are a number of interesting things happening now which I think are undertheorized:
What is the real economics of AI, given that the supply chains are so long and complex, consistent of both material and intellectual inputs, and the market for demand is uncertain? This is the trillion dollar question in terms of valuations, and it’s unanswered. The empirics here are not very good because things are far out of equlibrium.
Put another way: what does AI mean for the relationships between capital, corporations, labor, and consumers? Some of these relationships are mediated by rules about corporate law, intellectual property and data use, and so are determinable by law rather than technology. Information law therefore is a key point of political intervention in an economic system that is otherwise determined by laws of nature (energy, computation, etc.?
To put it another way: superintelligence has been happening and continues to happen. Some of this is due to laws of nature. But there is still a meaningful point of human intervention, which is the laws of humanity. Designing and implementing those laws well remains an important challenge.
One last thought. I’ve been inspired by Beninger’s The Control Revolution (1986) which is a historical account of the information economy in terms of cybernetics and information theory. You can ask an AI to tell you more about it, but one item comes to mind: that each new information technology first seems to threaten the jobs of people doing information work, and then leads to an expanded number of information jobs. This has to do with the way complexity is and is not managed by the technology. There’s an open question whether this generation of AI is any different. The question is truly open, but my hunch at the moment is that today’s AI systems are creating a lot more complexity than they are controlling. We will see.
One thing that’s become clear to me lately is that I now believe that ethics can be naturalized. I also believe that there is in fact a form of ‘natural law’. By this I mean that that there are rights and values that are inherent to human nature. Real legal systems can either lie up to natural law, or not.
This is not the only position that it’s possible to take on these topics.
One different position, that I do not have, is that ethics depends on the supernatural. I bring this up because religion is once again very politically salient in the United States. Abrahamic religions ground ethics and morality in a covenant between humans and a supernatural God. Divine power authorizes the ethical code. In some cases this is explicitly stated law, in others it is a set of principles. Beyond divine articulation, this position maintains that ethics are supernaturally enforced through reward and punishment. I don’t think this is how things work.
Another position I don’t have is that there is that ethics are opinion or cultural construction, full stop. Certainly there’s a wide diversity of opinions on ethics and cultural attitudes. Legal systems vary from place to place. This diversity is sometimes used as evidence that there aren’t truths about ethics or law to be had. But that is, taken alone, a silly argument. Lots of people and legal systems are simply wrong. Moreover, moral and ethical truths can take contingency and variety into account, and they probably should. It can be true that laws should be well-adapted to some otherwise arbitrary social expectations or material conditions. And so on.
There has historically been hemming and hawing about the fact/value dichotomy. If there’s no supernatural guarantor of ethics, is the natural world sufficient to produce values beyond our animal passions? This increasingly feels like an argument from a previous century. Adequate solutions to this problem have been offered by philosophers over time. They tend to involve some form of rational or reflective process, and aggregation over the needs and opinions of people in heterogeneous circumstances. Habermas comes to mind as a one of the synthesizers of a new definition of naturalized law and ethics.
For some reason, I’ve encountered so much resistance to this form of ethical or moral realism over the years. But looking back on it, I can’t recall a convincing argument for it. I can recall many claims that the idea of ethical and moral truth are somehow politically dangerous, but that this not the same thing.
There is something teleological about most viable definitions of naturalized ethics and natural law. They are would would hypothetically be decided on by interlocutors in an idealized but not yet realized circumstance. A corollary to my position is that ethical and moral facts exist, but many have not yet been discovered. A scientific process is needed to find them. This process is necessarily a social scientific process, since ethical and moral truths are truths about social systems and how they work.
It would be very fortunate, I think, if some academic department, discipline, or research institution were to take up my position. At present, we seem to have a few different political positions available to us in the United States:
A conservative rejection of the university of being insufficiently moral because of its abandonment of God
A postmodern rejection of ethical and moral truths that relativizes everything
A positivist rejection of normativity as the object of social science because of the fact/value dichotomy
Politicized disciplines that presume a political agenda and then perform research aligned with that agenda
Explicitly normative disciplines that are discursive and humanistic but not inclined towards rigorous analysis of the salient natural facts
None of these is conducive to a scientific study of what ethics and morals should be. There are exceptions, of course, and many brilliant people in many corners who make great contributions towards this goal. But they seem scattered at the margins of the various disciplines, rather than consolidated into a thriving body of intellect. At a moment where we see profound improvements (yes, improvements!) in our capacity for reasoning and scientific exploration, why hasn’t something like this emerged? It would be an improvement over the status quo.
Complex systems theory is a way of thinking about systems with many interacting parts and functions. It draws on physics and the science of modeling dynamic systems. It’s a trans-disciplinary, quantitative science of everything. It often and increasingly gets applied to social systems, often through the methods of agent-based modeling (ABM). ABM has a long history in computational sociology. More recently, it has made inroads into economics and finance (Axtell and Farmer, 2022). That’s important intellectual territory to win over because, of course, it is vitally important for both private and public interests. Its progress there is gradual but steady. ABM and complex systems methods have no dogma besides mathematical and computational essentials. Their eventual triumph is more or less assured. As I’ve argued, ABM and complex systems theory are thus an exciting frontier for legal theory (Benthall and Strandburg, 2021). For these reasons, one line of my research is involved in developing computational frameworks (i.e. software libraries, mathematical scaffolding) for computational social scientific modeling.
Contextual Integrity (CI) is an ethical theory developed by Helen Nissenbaum. It is especially applicable to questions of the ethics of information technology and computation. Central to the theory is the idea of “appropriate information flow”, or flows of (personal) information which conform with “information norms”. According to CI, information norms are legitimized by a balance of societal values, contextual purposes, and individual ends. The work of the CI ethicist is to wrestle with the alignments and contradictions between these alignments, purposes, and ends to identify the most legitimate norms for a given context,. When the legitimate norms are identified, it is then in principle possible to design and deploy technology in accordance with these norms.
CI is a philosophy grounded in social theory. It has never been robustly quantified and many people think this is impossible to do. I’m not among these people. In fact, much of my work is about trying to quantify or model CI. It should come as no surprise, then, that I now see CI in terms of complexity theory. It has struck me recently that what this amounts to, more or less, is a computational social theory of ethics! This idea is exciting to me, and one day I’ll want to write it down in detail. For now, I have some nice diagrams and notes from a recent presentation I wanted to share.
CI is a theory of ethics that is ultimately concerned with the way that values, purposes, and ends legitimize socially understood practices. The ethicist’s job, for CI, is to design legitimate institutions. A problem for the ethicist is that some institutions can be legitimate, but utopian, in that they are not stable behavioral patterns for the sociotechnical system. Complex systems theory, as a descriptive science, is well adapted to modeling systems and identifying the regular behaviors within them, under varying conditions. Borrowing a notion from physics, a system can exhibit many regular behavioral states, which we might call phases. For example, it is well known that water has many different phases, depending on the temperature: ice, the liquid water, steam, etc.
Norms have both descriptive and (ahem) normative dimensions. (This confusing jargon is part of why it’s so hard to make progress in this area.) In other words, for there to be an actually existing norm, it has to be both regular and, to be ethical according to CI, legitimate.
There are critics of CI who argue that one problem with it is that it assumes an apolitical consensus of information norms without addressing how norms might be distorted by, e.g., power in society. This is not terribly fair to Nissenbaum’s broader corpus of work, which certainly acknowledges political complexity (see for example the recent Nissenbaum, 2024). Suffice it to say here that not all individual end up being ‘legitimized’ when ethicists assess things, and that legitimization is always political. Moreover, individual ends and politics can, of course, often be the driver of system behavior away from legitimate institutions. We can’t always have nice things.
Nevertheless, it remains useful to consider how and under what conditions a system could remain legitimate despite technological change. This is what the original CI design heuristic is: a procedure for evaluating what to do when a new technology creates a disruptive change in societal information flows.
Ideally, for CI, when a new technology destabilizes the sociotechnical system’s behavior and threatens it with illegitimate practices, society reacts (through journalism, through ethics, through a political process, through private choices and actions, etc.) and returns the system to a regular behavioral pattern that is legitimate. This might not be the same behavior as the system started with. It might be even better. And that’s OK.
What’s bad, for CI, is if the system gets stuck in an illegitimate but still robust phase.
While there are some applications of CI that serve anodyne ends of parsing and implementing uncontroversial privacy rules, there are other uses of CI as a radical critique of the status quo. This is well exemplified by Ido Sivan-Sevilla et al.’s comments on the FTC ANPR on Commercial Surveillance and Lax Data Security Practices (2022), which is a succinct and to-the-point condemnation of the “notice and consent” practices in commercial surveillance. We live in a world in which standard, even ubiquitous, technology norms depend on “laughable legal fictions” such as the idea that users of web services are legitimate parties to contracts with vendors. It is well documented how these fictions have been enshrined into law by decades of pressure by the technology sector in courts and government (Cohen, 2019).
Together, CI and complex systems theory can show how society can be a winner, or loser, beyond the sum of individual outcomes. There are certainly those that have argued that, essentially, “there is no such thing as society”, and that voluntary, binary transactions between parties are all there is. An anarchic, libertarian, or laissez-faire system certainly serves the individual ends of some, and is to some extent stable until the lords of anarchy create new systems of rules that are in their interest. It is difficult to analyze the social costs of these political changes in terms of “individual harms”, because the true marginal cost is not measurable at the level of the individual, but rather at the level of the phase transition. A complex systems theory allows for this broader view of what is at stake.
This approach also, I think, helps convey the fragility of legitimate institutions. Nothing guarantees legitimacy. Legitimate institutions typically constrain the behavior of some actors in ways that they individually do not enjoy. There are social processes which can steer a system towards a more legitimate phase, but these will meet with resistance, sometimes fail, and can be coopted by bad faith actors serving their own ends.
Indeed, there are those who would say we do not live in a legitimate system and have not lived in one for a long time. “Legitimate for whom?” Even if this is so, CI invites us to have a productive dialog about what legitimacy would entail, by sorting out different motivations and looking at the options for balancing them out. This good faith search for resolutions is often thankless and unrewarded, but certainly we would be worse off without it. On the other hand, arguments about legitimate institutions that are divorced from realistic understandings of sociotechnical processes are easily deployed as propaganda and ideology to cover illegitimate behavior. Ethics requires a science of sociotechnical systems; sociotechnical systems are complex; complex systems theory is a solid foundation for such a science.
References
Axtell, R. L., & Farmer, J. D. (2022). Agent-based modeling in economics and finance: Past, present, and future. Journal of Economic Literature, 1-101.
Benthall, S., & Strandburg, K. J. (2021). Agent-based modeling as a legal theory tool. Frontiers in Physics, 9, 666386.
Cohen, J. E. (2019). Between truth and power. Oxford University Press.
Nissenbaum, H. (2024). AI Safety: A Poisoned Chalice?. IEEE Security & Privacy, 22(2), 94-96.
Through a few new emerging projects and opportunities, I’ve had reason to circle back to the topic of Artificial Intelligence and ethics. I wanted to jot down a few notes as some recent reading and conversations have been clarifying some ideas here.
In my work with Jake Goldenfein on this topic (published 2021), we framed the ethical problem of AI in terms of its challenge to liberalism, which we characterize in terms of individual rights (namely, property and privacy rights), a theory of why the free public market makes the guarantees of these rights sufficient for many social goods, and a more recent progressive or egalitarian tendency. We then discuss how AI technologies challenge liberalism and require us to think about post-liberal configurations of society and computation.
A natural reaction to this paper, especially given the political climate in the United States, is “aren’t the alternatives to liberalism even worse?” and it’s true that we do not in that paper outline an alternative to liberalism which a world with AI might aspire to.
John Mearsheimer’s The Great Delusion: Liberal Dreams and International Realities (2018) is a clearly written treatise on political theory. Mearsheimer rose to infamy in 2022 after the Russian invasion of Ukraine because of widely circulated videos of a lecture in 2015 in which he argued that the fault for Russia’s invasion of Crimea in 2014 was due to U.S. foreign policy. It is because of that infamy that I’ve decided to read The Great Delusion, which was a Financial Times Best Book of 2018. The Financial Times editorials have since turned on Mearsheimer. We’ll see what they say about him in another four years. However politically unpopular he may be, I found his points interesting and have decided to look at his more scholarly work. I have not been disappointed, and find that he clearly articulates political philosophy I will use these articulations. I won’t analyze his international relations theory here.
Putting Mearsheimer’s international relations theories entirely aside for now, I’ve been pleased to find The Great Delusion to be a thorough treatise on political theory, and it goes to lengths in Chapter 3 to describe liberalism as a political theory (which will be its target). Mearsheimer distinguished between four different political ideologies, citing many of their key intellectual proponents.
Modus vivendi liberalism. (Locke, Smith, Hayek) A theory committed to individual negative rights, such as private property and privacy, against the impositions by the state. The state should be minimal, a “night watchman”. This can involve skepticism about the ability of reason to achieve consensus about the nature of the good life; political toleration of differences is implied by the guarantee of negative rights.
Progressive liberalism. (Rawls) A theory committed to individual rights, including both negative rights and positive rights, which can be in tension. An example positive right is equal opportunity, which requires interventions by the state in order to guarantee. So the state must play a stronger role. Progressive liberalism involves more faith in reason to achieve consensus about the good life, as progressivism is a positive moral view imposed on others.
Utilitarianism. (Bentham, Mill) A theory committed to the greatest happiness for the greatest number. Not committed to individual rights, and therefore not a liberalism per se. Utilitarian analysis can argue for tradeoffs of rights to achieve greater happiness, and is collectivist, not in individualist, in the sense that it is concerned with utility in aggregate.
Liberal idealism. (Hobson, Dewey) A theory committed to the realization of an ideal society as an organic unity of functioning subsystem. Not committed to individual rights primarily, so not a liberalism, though individual rights can be justified on ideal grounds. Influenced by Hegelian views about the unity of the state. Sometimes connected to a positive view of nationalism.
This is a highly useful breakdown of ideas, which we can bring back to discussions of AI ethics.
Jake Goldenfein and I wrote about ‘liberalism’ in a way that, I’m glad to say, is consistent with Mearsheimer. We too identity right- and left- wing strands of liberalism. I believe our argument about AI’s challenge to liberal assumptions still holds water.
Utilitarianism is the foundation of one of the most prominent versions of AI ethics today: Effective Altruism. Much has been written about Effective Altruism and its relationship to AI Safety research. I have expressed some thoughts. Suffice it to say here that there is a utilitarian argument that ‘ethics’ should be about prioritizing the prevention of existential risk to humanity, because existential catastrophe would prevent the high-utility outcome of humanity-as-joyous-galaxy-colonizers. AI is seen, for various reasons, to be a potential source of catastrophic risk, and so AI ethics is about preventing these outcomes. Not everybody agrees with this view.
For now, it’s worth mentioning that there is a connection between liberalism and utilitarianism through theories of economics. While some liberals are committed to individual rights for their own sake, or because of negative views about the possibility of rational agreement about more positive political claims, others have argued that negative rights and lack of government intervention lead to better collective outcomes. Neoclassical economics has produced theories and ‘proofs’ to this effect, which rely on mathematical utility theory, which is a successor to philosophical utilitarianism in some respects.
It is also the case that a great deal of AI technology and technical practice is oriented around the vaguely utilitarian goals of ‘utility maximization’, though this is more about the mathematical operationalization of instrumental reason and less about a social commitment to utility as a political goal. AI practice and neoclassical economics are quite aligned in this way. If I were to put the point precisely, I’d say that the reality of AI, by exposing bounded rationality and its role in society, shows that arguments that negative rights are sufficient for utility-maximizing outcomes are naive, and so are a disappointment for liberals.
I was pleased that Mearsheimer brought up what he calls ‘liberal idealism’ in his book, despite it being perhaps a digression from his broader points. I have wondered how to place my own work, which draws heavily on Helen Nissenbaum’s theory of Contextual Integrity (CI), which is heavily influenced by the work of Michael Walzer. CI is based on a view of a society composed of separable spheres, which distinct functions and internally meaningful social goods, which should not be directly exchanged or compared. Walzer has been called a communitarian. I suggest that CI might be best seen as a variation of liberal idealism, in that it orients ethics towards a view of society as an idealized organic unity.
If the present reality of AI is so disappointing, then we must try to imagine a better ideal, and work our way towards it. I’ve found myself reading more and more work, such as by Felix Adler and Alain Badiou, that advocate for the need for an ideal model of society. What we currently are missing is a good computational model of such a society which could do for idealism what neoclassical economics did for liberalism. Which is, namely, to create a blueprint for a policy and science of its realization. If we were to apply AI to the problem of ethics, it would be good to use it this way.
In almost any contemporary discussion of politics, today’s experts will tell you that metaphysics is irrelevant.
This is because we are discouraged today from taking a truly totalizing perspective–meaning, a perspective that attempts to comprehend the totality of what’s going on.
Academic work on politics is specialized. It focuses on a specific phenomenon, or issue, or site. This is partly due to the limits of what it is possible to work on responsibly. It is also partly due to the limitations of agency. A grander view of politics isn’t useful for any particular agent; they need only the perspective that best serves them. Blind spots are necessary for agency.
But universalist metaphysics is important for politics precisely because if there is a telos to politics, it is peace, and peace is a condition of the totality.
And while a situated agent may have no need for metaphysics because they are content with the ontology that suits them, situated agents cannot alone make any guarantees of peace.
In order for an agent to act effectively in the interest of total societal conditions, they require an ontology which is not confined by their situation, which will encode those habits of thought necessary for maintaining their situation as such.
What motivates the study of metaphysics then? A motivation is that it provides one with freedom from ones situation.
This freedom is a political accomplishment, and it also has political effects.
The other day at BIDS I was working at my laptop when a rather wizardly looking man in a bicycle helmet asked me when The Hacker Within would be meeting. I recognized him from a chance conversation in an elevator after Anca Dragan’s ICBS talk the previous week. We had in that brief moment connected over the fact that none of the bearded men in the elevator had remembered to press the button for the ground floor. We had all been staring off into space before a young programmer with a thin mustache pointed out our error.
Engaging this amicable fellow, whom I will leave anonymous, the conversation turned naturally towards principles for life. I forget how we got onto the topic, but what I took away from the conversation was his advice: “Don’t turn your passion into your job. That’s like turning your lover into a wh***.”
Scholars in the School of Information are sometimes disparaging of the Data-Information-Knowledge-Wisdom hierarchy. Scholars, I’ve discovered, are frequently disparaging of ideas that are useful, intuitive, and pertinent to action. One cannot continue to play the Glass Bead Game if it has already been won any more than one can continue to be entertained by Tic Tac Toe once one has grasped its ineluctable logic.
We might wonder, as did Horkheimer, when the search and love of wisdom ceased to be the purpose of education. It may have come during the turn when philosophy was determined to be irrelevant, speculative or ungrounded. This perhaps coincided, in the United States, with McCarthyism. This is a question for the historians.
What is clear now is that philosophy per se is not longer considered relevant to scientific inquiry.
An ethnographer I know (who I will leave anonymous) told me the other day that the goal of Science and Technology Studies is to answer questions from philosophy of science with empirical observation. An admirable motivation for this is that philosophy of science should be grounded in the true practice of science, not in idle speculation about it. The ethnographic methods, through which observational social data is collected and then compellingly articulated, provide a kind of persuasiveness that for many far surpasses the persuasiveness of a priori logical argument, let alone authority.
And yet the authority of ethnographic writing depends always on the socially constructed role of the ethnographer, much like the authority of the physicist depends on their socially constructed role as physicists. I’d even argue that the dependence of ethnographic authority on social construction is greater than that of other kinds of scientific authority, as ethnography is so quintessentially an embedded social practice. A physicist or chemist or biologist at least in principle has nature to push back on their claims; a renegade natural scientist can as a last resort claim their authority through provision of a bomb or a cure. The mathematician or software engineer can test and verify their work through procedure. The ethnographer does not have these opportunities. Their writing will never be enough to convey the entirety of their experience. It is always partial evidence, a gesture at the unwritten.
This is not an accidental part of the ethnographic method. The practice of data anonymization, necessitated by the IRB and ethics, puts limitations on what can be said. These limitations are essential for building and maintaining the relationships of trust on which ethnographic data collection depends. The experiences of the ethnographer must always go far beyond what has been regulated as valid procedure. The information they have collected illicitly will, if they are skilled and wise, inform their judgment of what to write and what to leave out. The ethnographic text contains many layers of subtext that will be unknown to most readers. This is by design.
The philosophical text, in contrast, contains even less observational data. The text is abstracted from context. Only the logic is explicit. A naive reader will assume, then, that philosophy is a practice of logic chopping.
This is incorrect. My friend the ethnographer was correct: that ethnography is a way of answering philosophical questions empirically, through experience. However, what he missed is that philosophy is also a way of answering philosophical questions through experience. Just as in ethnographic writing, experience necessarily shapes the philosophical text. What is included, what is left out, what constellation in the cosmos of ideas is traced by the logic of the argument–these will be informed by experience, even if that experience is absent from the text itself.
One wonders: thus unhinged from empirical argument, how does a philosophical text become authoritative?
I’d offer the answer: it doesn’t. A philosophical text does not claim authority. That has been its method since Socrates.
Digifesto 