In the first chapters of his 1942 memoir “The World of Yesterday”, Austrian writer Stefan Zweig recounts his upbringing in late 19th century Vienna. He alternates between mourning the loss of the security and stability of pre-war bourgeois life in the Habsburg capital and the alienating effect that this relentless imposition of predictability and moderation in life had in the youth.
Even though Zweig describes the 1930s as a period of barbarism and erosion the foundations of European civilisation, a loss he equates to a thousand years of progress, he does commend the agency that the past decades have granted to the young:
“At school and at home today’s children are often allowed to express the preferences and wishes of their young, inquisitive heart openly – they are free, independent, natural creatures, while as soon as we entered the hated school building we had to keep our heads down (…) School, to us, meant compulsion, dreary boredom, a place where you had to absorb knowledge of subjects that did not seem worth knowing, sliced into neat portions” (p.52)
In this critique of the educational system within which he came of age, Zweig is putting his finger on a broader issue. It is the dilemma of whether to adopt the standardisation of flows of information into carefully curated bits, deemed optimal by an editorial authority of sorts. As such, this process becomes a double-edged sword: does it triage the information landscape to only provide that which is of quality? Or does it work as a means of homogenising thought in the ‘masses’, potentially driving conformity and political docility?
It seems to me that we live in times where a particular idea of progress has locked the narrative into a path towards a perfect personalisation of content curation. Pioneers have developed methods to handle the dizzying amount of information that is being produced and put into circulation second after second. From primitive World Wide Web’s site directories to Google’s keyword and ranking system, the consensus in what this progress looks like is that people need to quickly find what they are looking for.
In this narrative, algorithmic feeds bring a double element of progress. First, they process user behaviour in the form of inputs to ‘mould itself’ to that user’s habits and preferences when presenting results to their queries. Second, they attempt to predict what the user would be interested in consuming. Regarding this second factor, the novelty is that they attempt to cover the step that precedes the act of seeking information. Before any interpretation of a user’s specific query, algorithms exert influence on the kind of information you are looking for.
Algorithms, and ongoing innovation in AI, are advancing towards that ideal of individual personalisation. Compared to Zweig’s praise of interwar education, this paradigm follows a similar principle of respecting the individual’s agency, their “preferences and wishes” and their “inquisitive hearts”. It certainly seems to set its sights upon this. A content recommendation algorithm that adapts its decisions to user input would seem to be adapting to the particular preferences and wishes of an individual.
However, one thing does not necessarily lead to the other.
Zweig goes on describing the pre-war school system and calls it a cold apparatus of learning:
“But the very fact that it was taught according to a dry-as-dust plan made the lessons themselves dry and lifeless, a cold apparatus of learning that was never adjusted to the individual and, like an automaton programmed to recite the terms ‘good, satisfactory, unsatisfactory’, showed only how far we met the demands of the curriculum” (p.52)
Again, he describes it as a “cold apparatus of learning that was never adjusted to the individual”. In certain moments he resorts to the metaphor of moulding things to shape, of forcing a series of variously shaped minds into uniform, predesigned moulds.1
With this in mind, it can be argued that algorithmic curation has much more of a cold apparatus of learning quality than is publicly admitted.
Even if responding directly to individual input, there is no reason to not argue that algorithmic feeds or results are more of a rather arbitrarily adjusted impersonal mould than a malleable shape that adapts to the individual.
There is indeed an element of arbitrariness in them. This is due to the fallibility of the variables that they work upon, that is, what kinds of factors that supposedly reflect a personality they decide to operationalise. Yes, there is an input that is unique to each individual, and an algorithm that produces its output taking the particular characteristics of that input into account. But the output can end up simply being that: a modified mould whose shape is defined by echoes of our behaviour that in the end do not really relate to the inner-life that is supposed to drive these behaviours.
So algorithms can still be placed on the other side of the continuum between the impersonal supply of information and personalised curation. They can be effectively placed in the former end: an impersonal supply of information, a cold apparatus. But with it the problem of alienating forms of mass-scale models of information supply can be all the more slippery. They can emerge as a mirage in the road to perfect personalisation. Instead of the traditional model of a single, carefully curated cold apparatus for the masses, this new model consists of limitless collections of individual cold apparatuses, assigned to users according to flawed variables and providing rather obscure results.
So what are the implications of this line of thought?
First of all, we return to Zweig’s condemnation of the uniform, disciplining education of the late 19th century “Age of reason”. Granted, algorithms do not force users to sit and watch the content they provide for 5 hours a day as a teacher would in a school. However, easy access to smartphones and the internet in general, the ‘path of least resistance’ inclinations of our decision-making instincts, and the in-built, addictive nature of many services that become the core providers of information in many people’s lives, bring into question how much real freedom these gates to virtually limitless information they produce.
This topic is complex and has been extensively researched. But to put it shortly, the current algorithmic paradigm is also inseparable from the constant inflows of suggestive psychological nudges, developed habits, path dependencies, limited amounts of attention per day, the need for entertainment and so on. They all intertwine into a particular kind of limitation of individual freedom, where exercising sovereignty in our consumption of information to our best interests (ie. re-learning how to not depend upon feeds to decide on what to consume, developing media literacy, choosing our preferred information outlets and the amount of information we choose to digest per day) becomes increasingly harder as days pass. This is to say that in the culture and the technological affordances that we surround ourselves with, and the most accessible information supply and curation services out there, a default state consolidates where exercising this sovereignty requires a constant, deliberate, and frankly intricate process of preparation, habit-making and vigilance that most are not equipped to undertake.
Thus, the corollary here is that this algorithmic regime is in its own particular way a freedom-suppressing institution akin to the educational system critiqued by Zweig. The alienation and difficulty in accessing means of self-realisation that the school system created can also develop within the algorithmic regime. The great difference is that there is not a common apparatus of information that provides a communal sense of a shared experience regarding information (think for example content offered by newspapers and television that offer a sense of a shared reality and of current discussion topics).
Instead, each personal apparatus is its own, only shared insofar as it resembles in some ways another person’s current algorithmically decided settings. As a result, this previous unity in terms of what information is shared among people in a society severely fragments. In a way, there is a different impersonal education system for each individual. And the shared experience in this case is the medium itself, the algorithm’s mechanisms as they are reflected in the information that each user is served. Large groups of people then can only discuss the ‘behaviours’ that the algorithm’s outputs as a whole reflect instead of a common, or at least substantially overlapping, information ecosystems.
It is not worse nor better, both have their upsides and downsides. I personally lean towards the algorithmic regime. Yet I do think that it is far from ideal, and much work needs to be done in avoiding these mirages of perfect personalisation. It is necessary to seek and devise other ways of reaching a healthier relation with information. Information informs impressions, attitudes, decisions and relations, it is one of the most important factors to take into account for course correcting towards a better world.
- An ancient metaphor, already seen in the ancient Greek myth of Theseus and Procrustes’ bed. ↩︎
✵
Leave a Reply