For most of human history, controlling what people believed meant, among other things, controlling what they could read. The Spanish Inquisition burned books. The church suppressed heretical texts. Voltaire’s pamphlets went up in flames. If dangerous ideas could not circulate, you could control the truth.
But as literacy spread, something else spread with it. When more people could read, evaluate, and draw their own conclusions, outright suppression became harder to sustain. Rather than destroying information, those who wanted to shape belief found that flooding readers with more of it worked just as well.
The Nazis falsified massive amounts of economic data to hide the true cost of rearmament from international observers and from their own population (Tooze, 2006). Goebbels destroyed and manipulated the very source of reality as people understood it. The Soviet Union ran parallel statistical systems for decades, one for publication, one for internal use, because the gap between what the economy was producing and what it was supposed to produce had become too embarrassing to reconcile in a single set of numbers (Davies, 1994). Apartheid South Africa designed household surveys that structurally undercounted Black poverty; deliberate choices, as Posel (2001) documents, about what truth should matter and to whom.
A direct consequence of this is the practice that authoritarian governments often seek to subordinate institutions that produce independent data, such as statistical offices, universities and journalism. This is because if you control the instruments of verification and opposition of your data, you control what is true.
The book burning of 1933. American Photo Archive.
Democracy’s answer
Democracy was built, in part, as the institutional answer to exactly this problem. Its architecture assumes that power must always answer to someone, and that someone must always have access to independent information to hold it accountable. The organisations that historically served as referees are public health agencies, national statistics offices, international monitoring bodies, investigative journalism and so on.
Nowadays, that architecture is under pressure in ways that would have seemed extraordinary twenty years ago. The authority of these institutions is increasingly contested, their findings are treated as political artefacts and their funding has been cut. Many political movements now treat independent data institutions as partisan because the data is “wrong” or “fake”.
We have never had access to more information than we do today. And yet, across countries and contexts, public discourse feels less grounded in shared truth. To me, something in the abundance itself seems to be part of the problem.
The new distortion
As we explored authoritarian manipulation of data requires intent, someone deciding what truth should look like. But in the last decade something emerged alongside it, which needs no such decision, and it runs on incentives and profit.
Social media companies have access to behavioural data at a scale no public institution has ever approached. They know, with extraordinary precision, what you click on, how long you look, what you share, what brings you back. That data is rigorously analysed and continuously acted on, and the consistent finding, across platforms and populations, is that content provoking fear, outrage, and perceived threat generates more engagement than content that is calm, nuanced, or constructive (Brady et al., 2017).
Platforms did not set out to make the world feel like a dangerous place; the data told them what kept people on screen, and they built systems to produce more of it. The result is a represented world that feels more frightening, divided, and hostile than the one most people actually live in. The cause is less sinister than it sounds, not a conspiracy of executives wanting to cause harm, but something almost as troubling: measuring the wrong thing, and then optimising for it relentlessly.
It is worth being precise about what engagement actually means; a person who watches forty seconds of a violent video has generated a data point, and that data point accurately reflects what happened, but it tells us nothing about whether that person is better informed, more capable, or in any way, better served. Engagement is a specific behavioural response to a specific stimulus, optimised for a specific commercial purpose.
Data is a representation of reality
Behind all of this lies a simpler problem that I believe gets often overlooked. Data does not speak for itself; it is a representation of reality, shaped by a lot of choices, about what to measure, what to count, and whose experience makes it into the picture.
Consider a map of the sea. You study it carefully, hold it in your hands, you see its details, its precision, and it feels hard-won. Since you have never seen the sea yourself, you read everything about it, the temperatures, the tides, the wave patterns. One day, you feel more than ready to go and explore it.
You go to the coast for the first time. The wind hits you before you even see the water, sand gets into your shoes within the first ten steps, the salt dries your lips and when you finally stand at the shore, you feel the cold at your feet, the pull, the noise, the indifference of it all, and you realise that nothing had prepared you for this.
The map was accurate, and completely insufficient.
That gap shows up constantly in development practice and data. Any nutrition survey can tell you prevalence rates, but it cannot tell you that households skip meals on the days before a payment arrives, or that the woman reporting adequate intake is consistently eating last. That knowledge lives in the community, absent not because no one cared, but because the instruments we use are built for a different kind of precision than the problem actually demands.
What we lose
In development specifically, data is the medium through which cases are made, interventions designed, and choices justified. Living wage estimates, for instance, are built from household expenditure surveys, market price data, and nutritional modelling, each step involves contestable choices, and each output carries authority that depends entirely on the credibility of the institutions behind it. Weaken those institutions, and the numbers do not just become less reliable: the argument they were designed to support disappears with them.
The erosion of data institutions is therefore not an administrative inconvenience with limited real-world effect. When the organisations that build and interpret these instruments are defunded, politicised, or discredited, we lose the capacity to ask the right questions, and with it, the ability to hold power and injustice accountable. Supporting this trend we see that deregulation of social media compounds this. A statistical office that is defunded stops producing reliable numbers and platform that faces no accountability for amplifying outrage over accuracy keeps doing exactly that. Both are failures of the same kind, institutions that should protect our relationship with truth, left without the mandate to do so.
Data is not truth. It is the best instrument we have built so far to seek it, and it only works if the institutions behind it are independent enough to be trusted, and honest enough to show what they cannot see. A defunded statistical office stops asking the right questions, and an unregulated algorithm stops caring about the answers. The sea map you were reading gets thinner; however, the sea stays the same.
References
Brady, W. J., Wills, J. A., Jost, J. T., Tucker, J. A., & Van Bavel, J. J. (2017). Emotion shapes the diffusion of moralized content in social networks. Proceedings of the National Academy of Sciences, 114(28), 7313–7318. https://www.pnas.org/doi/10.1073/pnas.1618923114
Davies, R. W. (1994). Soviet economic development from Lenin to Khrushchev. https://archive.org/details/sovieteconomicde0000davi
Posel, D. (2001). What’s in a name? Race categorisation under apartheid and its afterlife. Transformation: Critical Perspectives on Southern Africa, 47, 50–74. https://transformationjournal.org.za/wp-content/uploads/2017/03/tran047005.pdf
Reddy, S. G., & Pogge, T. W. (2010). How not to count the poor. In S. Anand, P. Segal, & J. Stiglitz (Eds.), Debates on the measurement of global poverty (pp. 42–85). Oxford University Press. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=893159
Tooze, A. (2006). The wages of destruction: The making and breaking of the Nazi economy. Allen Lane. https://www.jstor.org/stable/40263929
Leave a Reply