Category Archives: science, technology and society

the worlds we can lose when intelligence becomes artificial

Leave a reply

In 1958, Hannah Arendt could see where were were headed:

This future man, whom the scientists tell us they will produce in no more than a hundred years, seems to be possessed by a rebellion against human existence as it has been given, a free gift from nowhere (secularly speaking), which he wishes to exchange, as it were, for something he has made himself. …

It would be as though our brain, which constitutes the physical, material condition of our thoughts, were unable to follow what we do, so that from now on we would indeed need artificial machines to do our thinking and speaking. If it should turn out to be true that knowledge (in the modern sense of know-how) and thought have parted company tor good, then we would indeed become the helpless slaves, not so much of our machines as of our know-how, thoughtless creatures at the mercy of every gadget which is technically possible, no matter how murderous it is. (Hannah Arendt, The Human Condition, 1958, p. 3)

What is “human existence as it has been given”?

For most of our history, most human beings have lived with other people whose names they know. They have worked individually and collaboratively with materials in their context to make an environment that I will call a “world.”

A world has these features:

  • It is imbued with moral significance, because other people have made it, given it meaning, cared about it, and been affected by it. An individual cannot interact with a world without causing good or harm to other people.
  • It is real, not imaginary, and therefore it is stubborn. It rarely turns out the way we want, but we can learn from experience to work more effectively with it.
  • The other people involved in any world hold partially conflicting interests and goals and can be stubborn in their own way. Both the materials and the people resist any single will.
  • Each person has partial and even biased knowledge, beliefs, and feelings about the world. But their varied ideas can accumulate as they express them and record them. Each person can therefore explore not only a world but the accumulated human experience of that world.
  • Because we must act in the company of other people and learn by acting, our “thinking and speaking” are closely connected.
  • Because our deepest concerns (moral, spiritual, and otherwise) relate to the world that we shape with our minds and hands, our “thought” is also connected to our “know-how.”
  • Each world typically predates each human being and survives the person’s death, yet each person can affect it. In fact, the birth of any human being automatically changes the world, if for no other reason than a birth turns people into parents, siblings, and other kinds of relatives.
  • There is not one world but many human worlds. But worlds can interact to various degrees without becoming subsumed into one bigger world.

Why it is good to live in a world

It is not obvious that living in this kind of world is the best imaginable form of life. Most people have envisioned heaven or a political utopia differently. (For instance, in an ideal world, the other people usually become less stubborn!) But I could make three arguments in favor of living in a world like this.

First, it seems plausible that homo sapiens evolved for such a life. Our brains, senses, and bodies are equipped to navigate it.

For instance, newborn infants already recognize faces, which are designed to communicate information and emotions. And our languages and cultures have accumulated deep resources for sharing a world with finite other human beings. The Proto-Indo-European language already used first-, second-, and third-person verbs and indicative, imperative, and subjunctive moods to make distinctions that are useful for group discussions about a common world. Thus a world is arguably our habitat.

Second, the combination of agency and humility seems morally compelling. It is fitting that we can affect our environment but not do just anything we individually want with it. And we should see our context as imbued with moral significance.

Third, navigating a world is a way for creatures like us to achieve comprehension, to make sense of matters. As Arendt writes:

There may be truths beyond speech, and they may be of great relevance to man in the
singular, that is, to man in so far as he is not a political being, whatever else he may be. Men in the plural, that is, men in so far as they live and move and act in this world, can experience meaningfulness only because they can talk with and make sense to each other and to themselves (1958, p. 4).

Threats to human worlds

Each human world has always been fragile, subject to destruction if invaders arrive, a plague strikes, or the community breaks down.

In addition, tyrants threaten any shared world because they can turn individuals into means to their solo ends.

Mass society puts each world at risk by bringing us into relationships with millions of others, whose names we will never learn. And mass economic exploitation makes matters worse. In Origins of Totalitarianism, Arendt says, “loneliness, on the experience of not belonging to the world at all, … is among the most radical and desperate experiences of man. [It is] is closely connected with uprootedness and superfluousness which have been the curse of modern masses since the beginning of the industrial revolution and have become acute with the rise of imperialism at the end of the last century and the break-down of political institutions and social traditions in our own time.”

When history seems to move quickly and beyond anyone’s control, humans cease to feel that they are agents in any recognizable world.

Ideology can be defined as any system of thought that substitutes core assumptions for actual engagement with other people in a common world.

Finally, although media can enrich any given world, it can also disrupt it. Imagine people sitting alone or in passive company before a TV screen that tells them about gruesome crimes. Their actual world may be safe, or less dangerous than it was in the past, but the mediated world is cruel.

New threats in the age of AI

This theoretical framework comes from Arendt, who drew on Heidegger’s fundamental insight that the human form of being (Dasein) is always “‘in’ the world in the sense that it deals with entities encountered within-the-world, and does so concernfully and with familiarity” (Being and Time, H105, trans. by Macquarrie & Robinson). Arendt makes Heidegger’s theory political and republican by emphasizing that people can talk and decide what to do with their worlds.

I have sketched this view to help make sense of a new phenomenon: intelligence that is artificial (AI). But Arendt already feared that we might “need artificial machines to do our thinking and speaking.”

When a person expresses a view, the content of what they say helps us to understand the world that the person inhabits. Even when people are flat-out wrong, the fact that they err or lie is part of our reality. In addition, a human view comes from a creature that can suffer. As such, it makes a claim on our compassion. In short, we attend not only to the content of the statement but also to the person who expressed it.

In contrast, when a large language model (LLM) answers a query (typically in the first-person singular and with emotive language like “I will be glad to …”), it does not reflect any particular perspective, nor does it come from a body that is capable of suffering. It just pretends to be a fellow participant in our world. We can attend to the words but not to the speaker.

Walter Cronkite was not really a visitor to Americans’ living rooms in 1970. He just appeared on TV screens. But he was a real person who could be assessed as such. An LLM is qualitatively different.

An LLM can be just another tool or resource, like a Heidegger’s hammer or perhaps like a library. I have collaborated with teams of Tufts engineering students to build the Civic Helpdesk and other applications of AI that are not yet publicly available. Working with them to fine-tune instructions or to design a user interface feels very much like collaborative work in a shared world. Note that I naturally said we “built” these tools, because the work feels roughly like building a shed, or perhaps an organization.

I have also developed what I think is a fairly tight practice of asking Claude about the Sanskrit and Pali original words in texts that I can only read in translation. This feels like a modest expansion of my inner life, if not a contribution to any shared world. (By the way, Claude is probably pulling these definitions from a finite set of published lexicons that have human authors.)

On the other hand, as Pope Leo notes in Magnifica humanitas, “current AI systems are more ‘cultivated’ than ‘built,’ for developers do not directly design every detail, but instead create a framework within which the intelligence ‘grows.’ As a result, fundamental scientific aspects — such as the internal representations and computational processes of these systems — remain, at present, unknown.” This sounds more like Arendt’s nightmare of a time when our thoughts cannot grasp what we have done.

The deepest concern is that we have developed biologically and culturally to flourish in what Arendt would call a world, but an individual who uses AI is no longer there.

See also: the papal encyclical on AI; Reading Arendt in Palo Alto; the human coordination involved in AI; the difference between human and artificial intelligence: relationships; the design choice to make ChatGPT sound like a human; and love of the world

the papal encyclical on AI

Leave a reply

Magnifica humanitas (“Magnificent Humanity”) is Pope Leo XIX’s XIV’s first encylical, subtitled “On Safeguarding the Human Person in the Time of Artificial Intelligence.”

Near the beginning, Leo makes a plea for “a shared discernment process.” He warns against worrying only about “contingencies” and “a succession of emergencies.” It is urgent to think ahout AI more deeply.

He observes that “most people are watching and waiting, observing from afar and merely hoping for the best. For this very reason, crucial questions impose themselves on our conscience and can no longer be avoided: Where are we going? Toward what goal do we wish to orient ourselves? What direction should we choose as a people and as a human community?”

When the Catholic Church practices “discernment” about social conditions, it “does not claim to offer ‘a definitive opinion'” but strives “to listen to and distinguish the many voices of our times and to interpret them in the light of God’s word.” Leo says that the Church is open not only to technical expertise but also to “a diversity of opinions” about values. Leo mentions previous papal letters with gratitude but also “gratefully acknowledges” the development of human rights doctrine through documents like the 1948 Universal Declaration of Human Rights, which did not originate with the Church.

Coming from outside the Church, I also welcome a basic inquiry into human intelligence at a time when computers are supporting a different kind of intelligence that poses risks for humans. The “central question” of the encyclical is indeed a basic question of our day: “what does it mean to safeguard our humanity?” I welcome the normative contributions of Catholic social doctrine, in much the way that Leo says he welcomes other views.

In paragraphs 11-14, Leo names four principles that are essential for “building a city founded on the common good.” The first is “a firm relationship with God.” I respect that idea but cannot follow it. But the second one is important and can be developed outside of Christianity. Leo says:

Today, the human desire for fullness of life is at risk of being misled by deceitful goals, such as the prospect of a technology that promises to free us from all weakness, and models of wellbeing that leave behind entire populations. All too often, we place our hope in unlimited ‘upgrades,’ in forms of progress that exacerbate inequalities, and in immediate solutions incapable of healing people’s wounds. As a result, while some pursue the illusion of unlimited self-assertion, many are deprived of basic necessities. The Church reminds us, with a firm yet humble voice, that true fulfilment is not achieved by eliminating weakness but through harmonious growth. It is found where freedom and responsibility are intertwined with mutual care and true solidarity, and where progress is measured by the dignity of each person and the good of all peoples.

This is very much like Hannah Arendt’s understanding of love for the world (amor mundi). We must love the species we happen to be (including the male portion, by the way). Our love for people should not be contingent on believing that we are good or smart.

When we can improve the human condition, we should. For example, if we can use gene therapy to cure a debilitating disease, then it is our obligation to do so. But the goal is never to perfect human beings. It is to help humans do the best we can with what we are, together.

Leo’s reference to a “city” based on love involves two Biblical stories that he briefly sketches near the outset. The Tower of Babel resembles a modern Large Language Model (LLM):

Fearing being scattered across the earth, [the people] sought to guarantee stability and power for themselves, and above all to ‘make a name’ for themselves. It was an impressive feat: a single language, a single technology, a single direction. However, the project concealed a profound danger. It was a project conceived without reference to God, supported by a uniformity that eliminated diversity and that chose homogenization over communion. When a city is built on pride and the claim to self-sufficiency, communication breaks down, languages are confused and people no longer understand each other. The result is not unity, but dispersion.

The other city is Jerusalem as it was rebuilt by Nehemiah–a story with deep civic resonance that I have discussed on this blog and in What Should We Do? A Theory of Civic Life (pp. 81-83)

The teachings of the Church are grounded in biblical narratives like these but have developed through previous efforts of discernment in modern times. Leo discerns the following components of today’s Catholic Social Doctrine: the equal dignity of all human beings; the supreme value of human rights; the principle of the common good; the principle of the universal destination of goods; the principle of subsidiarity; the principle of solidarity and the principle of social justice.

I will mention a few points that interested me from this framework.

First, the encyclical develops an interesting view of property rights in an era of data science and AI. Leo acknowledges that “certainly there is a right to private property, which has its own specific meaning and purpose.” (He does not explain the purpose of private property, but it could be to permit individuality and thus human dignity.) However, for Leo, private property is “always subordinate to the universal destination of goods,” which means that “the earth’s goods — soil, water, air and natural resources — are given by God to the entire human family to sustain the lives of all, and … every person has an inherent right to the use of such goods, both now and in the future.”

So far, this is established Catholic Social Doctrine, but the novel point in Magnifica humanitas involves intellectual property: “Today, among the goods that are universally intended for everyone, we must also include new forms of property, such as patents, algorithms, digital platforms, technological infrastructure and data.” Later, Leo says, “Data is the product of many contributors and should not be treated as something to be sold off or entrusted to a select few. It is necessary to think creatively in order to manage data as a common or shared good.”

Prevalent legal frameworks treat software and intellectual property as the work of human beings who have the right to own the fruits of their work. But Leo traces all goods back to God. When a human being invents software or a machine, there is no second creation. The output is still meant for the entire human family.

Leo also makes a strong argument against tech-bro economics:

It is important to ensure that this growth in appreciation of human dignity is not obscured by the pressure of new ideologies or very powerful interests in today’s world. Among these ideologies, I consider particularly insidious the one that suggests that every person must earn or justify his or her own worth, to the point of attributing greater value to those who are more efficient or effective. From this perspective, persons end up being reduced to a means of achieving results, a resource to be used and exploited, and are no longer recognized as a proper end in themselves who should never be instrumentalized. The value of persons, however, does not depend on what they achieve or produce. There are rights that apply to everyone simply by virtue of being human, and no human power can legitimately deny or arbitrarily limit them.

The practical concerns that the encyclical catalogues include propaganda and misinformation, loss of meaningful work, rising inequalities of power and wealth, and deadly militarism enabled by “autonomous” weapons systems.

These are valid topics, and the encyclical sometimes reads like a thoughtful but relatively conventional policy white paper.

At times, however, the specifically Catholic perspective lends additional depth to its conventional recommendations. For instance, here is some general advice regarding education in a time of AI:

We need adults to rediscover their vocation as artisans of education, prepared to work patiently each day, with the support of extensive and shared educational partnerships. Today, accompanying children and young people in using technology for developing responsible relationships, helping them to recognize the risks and choose what fosters inner freedom, is a concrete form of charity and will safeguard their dignity. Teaching new generations that technological evolution does not follow a predetermined path, but can be guided by personal and collective responsibility, constitutes one of the most valuable services to the common good.

A purely secular nonprofit could have written those sentences, although perhaps without the reference to charity (caritas). But only the Church would preface this passage with the preceding sentence: “Indeed, we must consider the digital world as a new continent to be evangelized, one that requires generous missionaries who are mature in the faith.”

I suppose I would have liked to read a bit more about the spiritual costs of intelligence that is artificial rather than an activity of the human mind. In the history of the Catholic Church, technology has repeatedly changed how human beings have formed and communicated ideas and meanings. The codex, the confessional, the cathedral, the printing press, and the broadcast studio have restructured individual and collective mentalities. We have survived such changes so far, as has the Church. But right now, we must think deeply and act effectively to prevent AI from reducing us to “data, a cog in a machine or a commodity” so that it can instead become an “instrument of growth, justice and fraternity.”

See also: Reading Arendt in Palo Alto; the Nehemiah story: on the pros and cons of walls; AI as Satanic; love of the world; the encyclical Laudato Si and the power of peoples to organize; etc.

what is a brute fact?

Leave a reply

During the twenties, so a story goes, [the former Prime Minister of France, Georges] Clemenceau, shortly before his death, found himself engaged in a friendly talk with a representative of the Weimar Republic on the question of guilt for the outbreak of the First World. War. “What, in your opinion,” Clemenceau was asked, “will future historians think of this troublesome and controversial issue?” He replied, “This I don’t know. But I know for certain that they will not say Belgium invaded Germany.” (Hannah Arendt, “Truth and Politics,” 1967, p. 239 )

Arendt uses this anecdote as an example of “brutally elementary data.” On p. 237, she mentions the “unyielding, blatant, unpersuasive stubbornness” of certain “truths seen and witnessed with the eyes of the body, and not the eyes of the mind.”

I agree that Belgium did not invade Germany in August 1914. (The reverse is true.) However, this example is complicated.

First, it is not a literal fact that “Germany” invaded “Belgium.” The name of any country is a concept, a metaphor, or a simplification. Perhaps the “brutally elementary data” is that some people moved from locations in German territory to locations in Belgian territory, and these people were (among other things) soldiers in the German Army. But even that formulation introduces information that would not be evident to an observer who was unaware of European politics.

Second, you and I do not remember seeing German troops cross the border. We believe that Germany invaded because that is what we have learned in school or from media. Our knowledge is entirely contingent on trust in these institutions.

Third, the word “invaded” is normatively loaded. An invasion isn’t necessarily bad. The Allied landings in Normandy were an invasion in a just cause. But Clemenceau uses the the word to imply that Germany broke its obligations and started the war. He would disagree with someone who said, “In August 1914, Imperial German troops had to extend the front into Belgian territory to protect the Fatherland,” even though that would also describe the same event.

Finally, Clemenceau used this example because he presumed–and expected his audience to presume–that the act of invading Belgium was the crucial causal factor. What if someone replied that the invasion was only one event in a sequence that begin with the assassination in Sarajevo on June 28, 1914, Austro-Hungary’s declaration of War on Serbia one month later, and Russia’s declaration of war against Austro-Hungary?

Clemenceau could have remarked, “They will not say that the Archduke Franz Ferdinand assassinated Gavrilo Princip.” (The reverse was the case). But he did not choose that example because his motive was to cast blame on Germany. There are infinite facts, and Clemenceau selected one to make a point.

Lenin argued that the cause of the First World War was imperialism. Europeans had run out of countries to conquer and exploit and had turned on each other. Some would say that Lenin’s thesis was an interpretation, whereas “Germany invaded Belgium” is a fact. But Clemenceau implied (or “implicated“) a whole interpretation by choosing a particular fact. And Lenin could cite many facts in support of his interpretation.

Insofar as we can know facts by direct observation or reliable methods, we don’t really need a variety of opinions to attain knowledge. If you think of a school, a university, or a newspaper as a purveyor of facts, then you may be uninterested in whether the people involved hold diverse views, and you may be suspicious when they seem to be editorializing. They should stick to the truth. Disagreement is a sign that an issue hasn’t yet been resolved–as it should be.

On the other hand, if you think that every important claim is an opinion, then you will see such institutions as forums for debate. (I think that is how Bari Weiss sees both CBS News and the University of Austin.) You may want these institutions to be pluralistic, but you won’t count on them to generate reliable information. And you may be quick to assert a right to disagree with any claim, no matter the nature of the evidence.

Presumably, we should navigate between these extremes, valuing both information and opinion and recognizing the two as intrinsically linked. Arendt wants us to remain connected to the actual world, and she is worried that ideology disconnects us from facts. But she also wants us to remain connected to other people, who inevitably have different interpretations. As she writes in The Human Condition (p. 57):

… the reality of the public realm relies on the simultaneous presence of innumerable perspectives and aspects in which the common world presents itself and for which no common measurement or denominator can ever be devised. For though the common world is the common meeting ground of all, those who are present have different locations in it, and the location of one can no more coincide with the location of another than the location of two objects. Being seen and being heard by others derive their significance from the fact that everybody sees and hears from a different position. This is the meaning of public life, compared to which even the richest and most satisfying family life can offer only the prolongation or multiplication of one’s own position with its attend ing aspects and perspectives. ….Only where things can be seen by many in a variety of aspects without chang ing their identity, so that those who are gathered around them know they see sameness in utter diversity, can worldly reality truly and reliably appear.

See also: ideological pluralism as an antidote to cliche; the case for viewpoint diversity; is all truth scientific truth?; holding two ideas at once: the attack on universities is authoritarian, and viewpoint diversity is important etc.

can AI solve “wicked problems”?

Leave a reply

I’ve been reading predictions that artificial intelligence will wipe out swaths of jobs–see Josh Tyrangiel in The Atlantic or Jan Tegze. Meanwhile, this week, I’m teaching Rittel & Webber (1973), the classic article that coined the phrase “wicked problems.” I started to wonder whether AI can ever resolve wicked problems. If not, the best way to find an interesting job in the near future may be to specialize in wicked problems. (Take my public policy course!)

According to Rittel & Webber, wicked problems have the following features:

  1. They have no definitive formulation.
  2. There is no stopping rule, no way to declare that the issue is done.
  3. Choices are not true or false, but good or bad.
  4. There is no way to test the chosen solution (immediate or ultimate).
  5. It is impossible, or unethical, to experiment.
  6. There is no list of all possible solutions.
  7. Since each problem is unique, inductive reasoning can’t work.
  8. Each problem is a symptom of another one.
  9. You can choose the explanations, and they affect your proposals.
  10. You have no “No right to be wrong.” (You are affecting other people, not just yourself. And the results are irreversible.)

Rittel and Webber argue that those features of wicked problems deflate the 20th-century ideal of a “planning system” that could be automated:

Many now have an image of how an idealized planning system would function. It is being seen as an on-going, cybernetic process of governance, incorporating systematic procedures for continuously searching out goals; identifying problems; forecasting uncontrollable contextual changes; inventing alternative strategies, tactics, and time-sequenced actions; stimulating alternative and plausible action sets and their consequences; evaluating alternatively forecasted outcomes; statistically monitoring those conditions of the publics and of systems that are judged to be germane; feeding back information to the simulation and decision channels so that errors can be corrected–all in a simultaneously functioning governing process. That set of steps is familiar to all of us, for it comprises what is by now the modern-classical mode planning. And yet we all know that such a planning system is unattainable, even as we seek more closely to approximate it. It is even questionable whether such a planning system is desirable (p. 159)

Here they describe planning systems that would have been very labor-intensive in 1973, but many people today imagine that this is how AI works, or will work.

why are problems wicked?

Some of the 10 reasons that some problems are “wicked,” according to Rittel & Webber, relate to the difficulty of generating knowledge. Policy problems involve specific things that have many features or aspects and that relate to many other specific things. For example, a given school system has a vast and unique set of characteristics and is connected by causes and effects to other systems and parts of society. These qualities make a school system difficult to study in conventional, scientific ways. However, could a massive LLM resolve that problem by modeling a wide swath of the society?

Another reason that problems are wicked is that they involve moral choices. In a policy debate, the question is not what would happen if we did something but what should happen. When I asked ChatGPT whether AI will be able to resolve wicked problems, it told me no, because wicked problems “are value-laden.” It added, “AI can optimize for values, but it cannot choose them in a legitimate way. Deciding whose values count, how to weigh them, and when to revise them is a normative, political act, not a computational one.”

Claude was less explicit about this point but emphasized that “stakeholders can’t even agree on what the problem actually is.” Therefore, an AI agent cannot supply a definitive answer.

A third source of the difficulty of wicked problems involves responsibility and legitimacy. In their responses to my question, both ChatGPT and Claude implied that AI models should not resolve wicked problems because they don’t have the right or the standing to do so.

what’s our underlying theory of decision-making?

Here are three rival views of how people decide value questions:

First, perhaps we are creatures who happen to want some things and abhor other things. We experience policies and their outcomes with pleasure, pain, or other emotions. It is better for us to get what we want–because of our feelings. Since an AI agent doesn’t feel anything, it can’t really want anything; and if it says it does, we shouldn’t care. Since we disagree about what we want, we must decide collectively and not offload the decision onto a computer.

Some problems with this view: People may want very bad things–should their preferences count? If we just happen to want various things, is there any better way to make decisions than to maximize as many subjective preferences as possible? Couldn’t a computer do that? But would the world be better if we did maximize subjective preferences?

In any case, you are not going to find a job making value-judgments. Today, lots of people are paid to make decisions, but only because they are assumed to know things. Nobody will pay for preferences. Life works the other way around: you have to pay to get your preferences satisfied.

Second, perhaps value questions have right and wrong answers. A candidate for the right answer would be utilitarianism: maximize the total amount of welfare. Maybe this rule needs constraints, or we should use a different rule. Regardless, it would be possible for a computer to calculate what is best for us. In fact, a machine can be less biased than humans.

Some problems with this view: We haven’t resolved the debate about which algorithm-like method should be used to decide what is right. Furthermore, I and others doubt that good moral reasoning is algorithmic. For one thing, it appears to be “holistic” in the specific sense that the unit of assessment is a whole object (such as a school or a market), not separate variables.

Third, perhaps all moral opinions are strictly subjective, including the opinion that we should maximize the satisfaction of everyone’s subjective opinions. Then it doesn’t matter what we do. We could outsource decisions to a computer, or just roll a die.

The problem with this view: It certainly does matter what we do. If not, we might as well pack it in.

AI as a social institution

I am still tentatively using the following model. AI is not like a human brain; it is like a social institution. For instance, medicine aggregates vast amounts of information and huge numbers of decisions and generates findings and advice. A labor market similarly processes a vast number of preferences and decisions and yields wages and employment rates. These are familiar examples of entities that are much larger than any human being–and they can feel impersonal or even cruel–but they are composed of human inputs, rules, and some hardware.

Another interesting example: integrated assessment models (IAMs) for predicting the global impact of carbon emissions and the costs and benefits of proposed remedies. These models have developed collaboratively and cumulatively for half a century. They take in thousands of peer-reviewed findings about specific processes (deforestation in Brazil, tax credits in Germany) and integrate them mathematically. No human being can understand even a tiny proportion of the data, methods, and instruments that generate the IAMs as a whole. But an IAM is a human product.

A large language model (LLM) is similar. At a first approximation, it is a machine that takes in lots of human generated text, processes it according to rules, and generates new text. Just the same could be said of science or law. This description actually understates the involvement of humans, because we do not merely produce the text that the LLM processes to generate output. We also conceive the idea of an LLM, write the software, build the hardware, construct the data centers, manage the power plants, pour the cement, and otherwise work to make the LLM.

If this is the case, then a given AI agent is not fundamentally different from a given social institution, such as a scientific discipline, a market, a body of law, or a democracy. Like these other institutions, it can address complexity, uncertainty, and disagreements about values. We will be able to ask it for answers to wicked problems. If current LLMs like ChatGPT and Claude refuse to provide such answers, it is because their authors have chosen–so far–to tell them not to.

However, AI’s rules are different from those in law, democracy, or science. I am biased to think that its rules are worse, although that could be contested. The threat is that AI will start to generate answers to wicked problems, and we will accept its answers because our own responses are not definitively better and because it responds instantly at low cost. But then we will lose not only the vast array of jobs that involve decision-making but also the intrinsic value of being decision-makers.

Source: Rittel, Horst WJ, and Melvin M. Webber. “Dilemmas in a general theory of planning.” Policy sciences 4.2 (1973): 155-169. See also: the human coordination involved in AI; the difference between human and artificial intelligence: relationships; the age of cybernetics; choosing models that illuminate issues–on the logic of abduction in the social sciences and policy

why policy debates continue

Leave a reply

I’m at Stanford today to discuss a paper, Policy Models as Networks of Beliefs. After circulating my draft, I realized that the following is really my argument. …

We use mental models to think about and discuss contested questions of policy. Worthy models typically have these features:

  1. They have many components, not just a few. A model might include a causal inference, such as “spending more on x produces better outcomes.” But those two components (the spending and the outcomes) must be part of a much larger model that also explains why certain outcomes are valuable, where the money would come from, what else effects the system, and so on.
  2. The components should be connected, and the resulting structure matters. Structures can take various forms (e.g., root-cause analysis, vicious cycles). There is no single best structure.
  3. Pieces of models may prove regular. For instance, maybe spending more on x regularly produces better outcomes, all else considered. But such regularities only apply to small aspects of good models. The science-like effort to find regularities can only get us so far.
  4. Some components of any worthy model should be values or normative claims. Some normative components have regular significance in all models. However, many value components change their significance depending on the context. Equality, for example, does not consistently mean the same thing and may not always be desirable.
  5. If a model proves influential, it can change the world, which can require a new model. For example, arguing that more money should be spent on X could cause more funds to be allocated to X, at which point it would no longer be wise to increase the funding. Models are dynamic in this sense.

I believe this account supports a pluralistic, polycentric, pragmatist, and deliberative approach to policymaking, as opposed to a positivistic one.

See also: choosing models that illuminate issues–on the logic of abduction in the social sciences and policy; different kinds of social models; social education as learning to improve models; etc.