Automating critical thinking
Can critical thinking be codified? To discuss this theme, three tools have to be used together: real software abstraction, abductive modelling and queer critique of normativity.
While it may be lucrative to think about right and wrong assumptions when looking at software and things it obscures and reveals, we need better terms to define what it is that we see as the weak side of scaling the complex epistemic infrastructures. In this letter, I try to consider this weakness as the failure of technology to think critically about itself, or more precisely, to evaluate its normative assumptions.
Real software abstraction
Right off the bat, we have to drop judging which assumptions are right or wrong because taking on any responsibility for moral judgement leads to the all too familiar dynamic of one agency being responsible for deciding which assumption content is more ‘right’ than the other. For example, where social media firms such as Meta or X are held morally responsible for the content that their assumptions carry, some other Web 2.0 industry giants, such as product tools like Atlassian Jira, Asana or ClickUp are not likely to ever find themselves in such situation. The difference might be best explored through the level of abstraction at which the assumptions are held in software systems - the ability to look beyond the assumptions together and think instead about the relations between them and the processes that create them. For such a shift in focus, we should think about ‘assumptions in the abstract’, - that is, to have the assumption to be present as an abstract entity unspecific to what it assumes. This is, in essence, what software scholar Robert W. Gehl, proposes when he speaks about real software abstraction. Writing in the early 2010s, he looks at the example of MySpace losing its position to the growing popularity of Facebook. To him, Facebook from the very start offered its users a better infrastructure for capturing and accessing knowledge, which had taken the social relation in an abstract form separately from the implementation of the system itself. For example, where MySpace had an exceedingly flexible interface and offered its users a ‘concrete chaos’, it overwhelmed the users who eventually opted for Facebook’s more unified - or abstract - way of dealing with the content. This kind of software abstraction for Gehl is real, in the Marxian sense, because it bears real effects on the social relations which the platform cultivates. (Gehl, 2014:18.)
Linking this back to what we discussed above, taking software assumptions in the abstract is a real software abstraction in the same sense because it splits the driving idea of the software system and the various contents it may have into symbolic forms. Once a symbol, the assumption can be epistemically referenced, stored, enumerated, deferred, displaced and otherwise computationally manipulated separately - and together with - other software processes, such as production routines, stakeholder negotiations or operations.
We may know what we don’t know in specific circumstances, yet we are not able to scale these ignorance considerations, essentially having to discover what we don’t know in each case.
Abduction as ignorance-preserving tactic
Thinking about assumptions in abstract terms works well for revealing the intentions of technologists and firms, yet it does not give us tools for automating the production of such abstractions. In other words, we may know what we don’t know in specific circumstances, yet we are not able to scale these ignorance considerations, essentially having to discover what we don’t know in each case. Here, the abductive logical move, as suggested by media theorist Clemens Apprich, seems the most appropriate solution. (Apprich, 2023).
Abduction, in the sense used by logician Charles S. Peirce, is a mode of reasoning which arrives at new ideas by temporarily suspending a disbelief or lack of knowledge about its premises. In this sense, abduction is different from the other two of Peirce’s reasoning modes, induction and deduction, in that the latter two arrive at the consequent through something which is already known for a fact. (Peirce, 1955: 197.) The abductive model, alternatively, allows the system to preserve the ignorance for the time being, and to act on the assumption until the situation changes in a way that proves or disproves the assumption. (Magnani, 2009: 65.) The ignorance preservation says, we don’t know X yet, but we know that there’s an X we don’t know. Now, if we assume that X is Y rather than W, then we can imagine where it would get us. Linking back to real software abstraction, abductive modelling works well with assumptions in the abstract, because the abductive move does not have a goal to discover what the black-boxed assumption was all about. Rather, it looks for a plausible enough assumption that would give the reasoning, either human or machinic, the shortest path to the outcomes the system aims for. In this sense, the abductive model is content-agnostic, as Facebook in Gehl’s example, - which of course does not mean that it is worse or better than any other, for example, MySpace-style models. Look at all the wrongs the Facebook abstraction got into over the years, including having to make difficult moral decisions! Yet, it is certainly more scalable.
What is required here is a mechanism that allows to distinguish the norm, and deploys abduction in a reverse way, to find out the nonnormative, other and deviant.
The meaning of queer
While preserving ignorance may allow the system to scale its epistemic infrastructures, the key question of how to automate critical thinking remains unanswered. The first thing that comes to mind is that critical thinking is something that works against the code and therefore cannot be codified. In other words, critical thinking is a kind of thinking that pays attention to all things that are unseen or invisible in the lens of the present episteme, or the system of knowledge. What is required here is a mechanism that allows to distinguish the norm, and deploys abduction in a reverse way, to find out the nonnormative, other and deviant. This mechanism is queer study. There are two distinct and slightly contradictory uses of the term ‘queer’, to follow the theorisation of Siobhan B. Somerville, both of which, however, deal with critique of normativity. In its first meaning, queer is an umbrella term for a range of sexual and gender identities which are straight or different from the norm. The second meaning is more broad and applies to our case of real software abstraction. In this meaning, queer deals with the production of normativity in any sense, exposing it to the intersection of difference and power. This opens the notion of ‘queer’ to a variety of social processes including race, indigeneity, gender, ability, class, religion or nation. (Somerville, 2020: 2.)
In this sense, adding the queer layer to the automated abductive mechanism of abstraction gives the system new tools for producing its epistemology. Now it is not only produced at scale in a content-agnostic way along the norms prescribed by the assumptions the system operates with. It is also queer in the sense that it has a built-in mechanism for detecting the normative trends and directing the abduction modelling processes in trajectories that explore the terrains located outside of the system’s normativity.
References
Apprich, Clemens. 2023. ‘AI Bayes – Bayesian Networks as Tools of Diagnosis’. https://enginesofdifference.org/2023/10/11/bayesian-knowledge/.
Gehl, Robert W. 2014. Reverse Engineering Social Media: Software, Culture, and Political Economy in New Media Capitalism. Philadelphia, Pennsylvania: Temple University Press.
Magnani, Lorenzo. 2009. Abductive Cognition: The Epistemological and Eco-Cognitive Dimensions of Hypothetical Reasoning. Vol. 3. Cognitive Systems Monographs. Berlin, Heidelberg: Springer. https://doi.org/10.1007/978-3-642-03631-6.
Peirce, Charles Sanders. (1940) 1955. Philosophical Writings of Peirce. Edited by Justus Buchler. New York: Dover Publications, Inc.
Somerville, Siobhan B., ed. 2020. The Cambridge Companion to Queer Studies. Cambridge ; New York, NY: Cambridge University Press.