AI, analogy, innovation, and commodity power in Spyros Samothrakis’ recent contributions

Spyros Samothrakis recently published an article (2024) and made a presentation in a series of virtual seminars organized by our research group[1]. His contributions are stimulating and deserve a serious discussion. I will try to participate in such a discussion by first underlining the aspects I think are promising in these contributions and then focusing on the issues I think are more problematic. I will conclude with a few ideas about overcoming the difficulties I see.

My understanding of Samothrakis’ proposal

Samothrakis starts his article (and, in even more detail, his presentation) with a pessimistic view of the current developments in AI research. Writing against the current media and industry discourse, Samothrakis tries to tone down the “overhyped rhetoric” (2024). However, his sobering comments are not (only) aimed at the AI community but much more at those who try to envision a new economic organization through models he dubs computerized economic planning (CEP). In Samothrakis’ contributions, CEP primarily refers to models and ideas inspired by Cockshott and Cottrell’s seminal book (1993). He, therefore, tries to moderate another hype, the one we saw in recent publications (Philips and Rozorwsky 2019), enthused by technological progress, that considers that socialist planning by computers has entered the “age of their feasibility” (Groos 2021).

Hence, Samothrakis does not work so much on how AI could be used to help the computerized planning of the economy; he uses more AI as an analogy that can help us understand what goal the CEP research community should try to achieve. The message is simple: lower your expectations. AI research has been confronted with two major problems: the generalized use of tacit knowledge by humans and the difficulty in discriminating which events are essential and which are secondary when trying to predict a future outcome. Here comes the analogy: Samothrakis postulates that CEP researchers will face (or are already facing) problems similar to those AI researchers already face. This analogy is only possible because Samothrakis uses the term planning in a different way than it has been used in economics (as in other social sciences and everyday parlance) for a dozen decades, but in the sense that it is used in the AI research community. This is not only semantics; what he means by “planning” refers to a completely different activity.

According to Samothrakis, what economists and social scientists call planning when talking about a planned economy should be called “value mixing.” Value mixing, for him, is allocating the correct number of resources and labour to the proper production process, aiming to produce a set quantity of goods and services and trying to reach pre-established goals. For Samothrakis, what should be called planning is the capacity to discover new ways of satisfying human needs or desires (these ways of satisfying human needs are called ‘concepts’) and establishing “whether or not a concept kiproduces any value and how to build new commodities” (Samothrakis 2024). His vision of planning covers part of what economists call innovation and part of what they call investment. Having a computer do this kind of planning is clearly different than asking it to resolve an input/output table (as in the Cockshott and Cottrell model) or to establish indicative prices based on iterative plans coming from councils (as in participatory economics, Albert and Hahnel 1991). It involves a capacity for discovery and prediction that brings the idea of planning to another level.

We then understand the value of the analogy. If we expect computers to be able to do this kind of planning, we will face problems similar to those faced by AI researchers. A lot of what we do and how we do it is something we know but cannot always explain (it is tacit knowledge), so coding it for a computer to understand turns out to be quite a complex task. Numerous events can change the course of the design or production of new ‘concepts’  and commodities (be it done by machines or by humans). The influence of these events is so unpredictable that knowing well enough what will happen in an economy for this knowledge to be useful for an algorithm to help ‘plan’ the economy (in the way Samothrakis means planning) is almost impossible.

Hence, the aforementioned sobering call Samothrakis makes about the potential of CEP. He suggests that planning should focus primarily on producing highly ‘powerful’ commodities. Power is used here in a very specific way: a commodity is powerful when it “maximises [one’s] future freedom of choice” (Samothrakis 2024). “An example of a powerful and rewarding commodity would be water, as it is necessary for life and can be used no matter what we want to make. A pre-packaged meal with duration of 1 day is a disempowering commodity, as it can only be used to be consumed (i.e. it has one use)” (Samothrakis 2024). Samothrakis’ main idea is thus that CEP could concentrate on organizing the planning of powerful goods while households and communities could assemble and modify them for the specific use they want to make of them. This would considerably reduce the quantity of planning needed and focus it on “concepts” of a more general use.

This fruitful idea leads him to suggest a model on which a planning institution could be based. He divides this institution into five departments:

  • Value identification: identify where value is created and capture it, in support of commodity power, decentralisation, and plasticity.
  • Value protection: prevent value from being destroyed, in support of commodity power, decentralisation, and stability.
  • Value mixing: help bring future value to the present, in support of reward maximisation, decentralisation, and equivalence of process.
  • Value ethics: distribute commodities fairly, in support of decentralisation, and also ensure that value is not extracted in ways that would be immoral.
  • Meta-value: evaluation of the programme. (Samothrakis 2024)

According to Samothrakis, this planning institution is not designed to replace capitalism as a system but to be integrated into a market economy and transform capitalist social relations in the longer term. In his article’s closing remarks, he also mentions starting small “settlements” to experiment with this institutional “setup.” To use Wright’s (2019) vocabulary, his proposal is not in rupture with the capitalist system but more of a symbiotic or interstitial nature.

My issues with Samothrakis’ proposal

As stimulating as it is, Samothrakis’ proposal leaves me uncertain about three significant aspects. I will present these issues and, in the next section, explain how I suggest they can be overcome.

My first question is about investment. Samothrakis offers a quite virulent critique of previous models, presenting them as if they only offered to plan a static economy with no innovation.

If we decide to use input–output tables given the quantities of an economy in 2023 (when this paper is written), all we are going to have in 2030, after quite a bit of effort, is a perfect 2023 economy. At best, all an optimal economy that is (naively) based on input—output tables can do is play catch-up with a neighbouring capitalist economy—it is unable to conjure new valuable concepts! As the capitalist economy continuously creates new concepts and new production methods, the socialist economy would try to create an optimal version of the capitalist past. (Samothrakis 2024)

A few lines after this quotation, he attacks the attempts of previous models to integrate innovation in their functioning as “often afterthoughts and responses to outside criticism” (Samothrakis 2024). Very well, but all this is going a bit too fast. First, not all previous models he quotes use input/output tables for the “static” (day-to-day production and consumption) economic aspect. For example, the model developed by Pat Devine (and, later, in collaboration with Fikret Adaman), called negotiated coordination, leaves this static aspect of the economy to the market. What is coordinated through negotiation is investment, and investment is a big part of what makes the dynamic aspect of an economy. New infrastructures, for sure, but also new R&D funding, new product lines, new services, all of this is decided upon through decisions about investment (and in the capitalist context, mainly using the criteria of return on investment). This is how an economy is transformed and transforms the world around it, for the better or worse (in our current economies, much for the worse, locking us in a rentability-seeking trap for decades to come, but that is for another discussion). Participatory economics and Cockshott and Cottrell’s model also have decision-making processes around investments. So they all think about innovation, but they approach it from a different point of view than Samothrakis does.

What is the main difference between these models and Samothrakis’ proposal? It seems like it primarily rests on their different visions of democratic planning; it can be seen either as a set of institutions inside a postcapitalist society or as following the AI analogy as Samothrakis does. All previous democratic economic planning models first analyzed the institutions of capitalist societies and what it meant to overcome them. Based on that analysis, these authors proposed a series of institutions to replace the ones stemming from capitalism. Potentiality for innovation is dispersed throughout these institutions, and innovation itself is seen as diversified (not only new end products but new production methods, work organization and ways of living). Depending on the model, it can come from a consumer request in a council, a new habit revealed in data, a production unit wanting to try something new, a project’s R&D committee suggesting a design or a negotiating body discussing investment plans. What previous models did is to try to understand where and how each institution could manage such new requests in a (more or less, depending on the model) decentralized way.

Following its AI analogy and adopting a reformist strategy, Samothrakis has an entirely different perspective. The institutions of production and consumption are not democratized; we are still in a society where the law of value governs production and investment decisions; the state is still a representative government, and plutocrats rule the world. CEP produces a planning institution that is integrated into the capitalist economy (Samothrakis does not tell us how this integration is made, but we get why; it would probably need a whole other paper to do this) but also reorganizes a large part of this economy: the planned production and distribution of powerful goods. Since it is only one institution integrated into the current capitalist set of institutions (market, power hierarchies, representative government, etc.), it needs to concentrate many functions that, in previous models, were distributed between different institutions. Even more so when following the AI analogy: if this institution is understood as an actor discovering the world around it and using the creation of commodity (or “concepts”) as its primary (or only?) way to interact with that world, then creating a department to do innovation (or “value identification”) is the logical answer to the premises of Samothrakis’ article.

However, as is always the case with reformist strategy, the question is: how far will it go? Or, to ask it the other way around: how integrated will it be in capitalism? If the responses to our problems are commodities (as powerful as they may be), if we aim at keeping the same consumption level and if we work in the same hierarchical organizations as we do today, it might be an impressive technical feat to plan the production of powerful commodities and our lives might be less alienated thanks to the DIY assemblage of commodities we will do in our living rooms instead of buying them at Walmart, but were does that lead us?

The question is even more daring when we consider the definition of democracy used by Samothrakis. Being rightfully critical of representative governments, he considers election a popularity contest and would instead use sortition to choose people working in his planning institution. Very well, but that misses the point. Democracy is not about electing people; it is about making decisions together. This is why the previous models rely so much on self-management at the micro level (especially participatory economics and negotiated coordination) and ancient Athenian-style direct democracy at the macro level (especially the Cockshott and Cottrell model that is central to Samothrakis’ discussion). Since democratic decision-making is mentioned nowhere in his contribution, a strange feeling emerges from Samothrakis’ contribution, as if the actual top-down hierarchical organization of society was not something it meant to address. For the person not staffed in the planning institution he has designed, it seems that life would look pretty much like today, with no more capacity to act autonomously on the organization of society.

What we can learn from Samothrakis’ proposal and how its problems could be overcome

After these critiques, I want to reiterate that Samothrakis’ article and presentation are stimulating contributions. His great strength is proposing and offering something, not just discussing the pros and cons of precedent models. In this spirit, I want to suggest how further research could use his contribution.

One thing Samothrakis did that I find inspiring (partly because it crosses some of my current reflections on postcapitalist economies) is that he designed an institution and not a complete model. He thinks this institution can be integrated into our current society, but I mostly see that it can be incorporated into many other postcapitalist models and reflections. I believe developing institutions or sets of institutions (what we in our research group started calling modules) is more promising than building models (and comparing, criticizing, and bettering them). I have already written at length on this so I won’t repeat it. Nevertheless, I think that Samothrakis, mostly in the second half of his article (about which he, unfortunately, did not talk much during his presentation), does exactly what I have in mind when I talk about building institutions and modules.

In this specific case, it is interesting to see that models based on a central agency for planning (the planning agency in Devine (1988) and Planning in Cockshott and Cottrell (1993)) have yet to propose an internal functioning of this agency. I believe most of Samothrakis’ proposals could be integrated into these models without significant modifications from one part or the other. This leads us to the discussion of chronology and transition (much easier to have when talking about institutions and modules than about models). The institution Samothrakis proposes, if implemented by a government – or based on an autonomous social economy structure, where this is possible – can be envisioned as a stepping stone or a transitional phase towards greater democratization of the economy. I can easily see how this could be used in a quasi-market-socialist context or by a revolutionary government building its population’s economic autonomy. The idea would be to have a “first step” of planning: some commodities being produced through this process and getting people to slowly build the non-market sector to reduce the population’s dependence on private corporations. This is my reinterpretation of Samothrakis’ proposal; I do not pretend it is faithful to the original, but I find it engaging.

That being said, for me, the most inspiring idea in Samothrakis’ proposal is about powerful commodities. I see it as the actualization of ideas already present in Murray Bookchin and Ernst Schumacher about human-scale technology, in Ivan Illich about conviviality, and in André Gorz about auto-production. Let’s synthesize briefly. Bookchin (1995) and Schumacher (1993) taught us that technology is not neutral; it transforms society to its image. Big and complex technology that only specialists can control produces a technocratic society, even more so when we depend on that technology. Illich (1973) showed that we should look for conviviality when considering technology. For him, conviviality means for humans to control their tools and for these tools to serve the people integrated into a community. Two decades later, André Gorz (2018) developed a vision of socialism where auto-production would be central. By that, he meant that we should spend a lot of our time working on things we use directly: growing a garden, building simple tools, taking care of our children and those of our neighbours, doing arts and crafts, etc. (an idea we find in William Morris’ writings too).

All this is fine, but unless I spend years learning tons of technical skills, I will never be able to build an electric mixer that will allow me to do more cooking at home. That’s where Samothrakis’ idea of powerful commodities is sound. If I can have access to the central engine of the mixer (that could also be used for other usages inside or outside of the kitchen) through a planned and centralized production and distribution process, the “add-ons” that I need to make it work as a mixer can be built by a smaller shop, closer to me that can offer a variety of local specificities. In the case of simple stuff, I can even create it myself or in my community’s close-by and simple infrastructure. This links well with the well-documented practices of the makerspaces and tools libraries, where people reinforce autonomy, creativity, and environmental sustainability by working together.

By giving us a glimpse of a usage of technology that is simultaneously centrally planned and community-based, Samothrakis has given us material for a yet-to-be-built consumption module.



Albert, Michael, et Robin Hahnel. 1991. The political economy of participatory economics. Princeton: Princeton University Press.

Bookchin, Murray. 1995. From Urbanization to Cities : Toward a New Politics of Citizenship. London; New York: Cassell.

Cockshott, W. Paul, et Allin Cottrell. 1993. Towards a new socialism. Nottingham, England: Spokesman.

Devine, Pat. 1988. Democracy and economic planning: the political economy of a self-governing society. In Aspects of Political Economy, Boulder, Colo: Westview Press.

Gorz, André. 2018. Ecologica. Kolkata: Seagull Books.

Groos, Jan. 2021. « Distributed Planned Economies in the Age of Their Technical Feasibility .» Behemoth 14 (2): 75‑87.

Illich, Ivan. 1973. Tools for Conviviality. New York: Harpers & Row.

Phillips, Leigh, et Michal Rozworski. 2019. The People’s Republic of Walmart: how the world’s biggest corporations are laying the foundation for socialism. London ; New York: Verso.

Samothrakis, Spyridon. 2024. « Artificial Intelligence and Modern Planned Economies: A Discussion on Methods and Institutions .» AI & SOCIETY, janvier.

Schumaker, Ernst. 1993. Small is beautiful. New York: Vintage Books.

Wright, Erik Olin. 2019. How to Be an Anticapitalist in the Twenty-First Century. New York: Verso.



[1] I would like to thank Sophie Elias-Pinsonnault and Jonathan Durand Folco for their comments on this logbook entry.

Belongs to the research project:
Democratic economic planning

Research axis :