For all latest news, follow the Daily Sun's Google News channel.
Imam Nahil’s Daily Sun’s 4 October article, “Attention Economy and Rising Social Polarisation’’ raises the right question for Bangladesh: should policy concentrate only on what people post, or also on the systems that decide what millions are encouraged to see? His case for algorithmic transparency, platform accountability and digital literacy is persuasive, and he is right that the answer is not greater government control over speech. This response adds the economic logic beneath it: why do commercially successful platforms have incentives to encourage engagement that is profitable for them but costly for society?
Where the Revenue Comes From
The attention economy is a rare example of a deeply rooted applied microeconomic issue: its core questions concern the incentives platforms face, what they seek to maximise, what users effectively “pay” with their attention, how advertising revenue is generated, and what happens when private returns diverge from social costs. For advertising-financed platforms, attention is the raw material. The longer users stay, the more advertisements can be shown, the more preference data collected, and the better recommendation systems predict what will hold attention next. Engagement, whether watching, clicking, sharing or commenting, is the measurable form of attention, so it is the quantity platforms are built to raise.
Nothing here requires a platform to prefer anger or falsehood. The incentive is simply to retain attention. But material provoking fear or indignation draws strong reactions, and a system designed to find what sustains engagement will tend to give it greater reach. The evidence Nahil cites points this way: a model in which weighting likes and shares raises engagement together with misinformation and polarisation, and a nine-country Twitter study finding polarised networks and toxicity towards out-groups. This does not show that social media creates division; it shows that commercial logic and social outcome are connected.
Private Return and Social Cost
Users generally pay no money price. The commercial exchange is between platform and advertiser, and what is sold is users' attention. Society, including people who never see a particular advertisement, is party to neither side, yet it lives with the by-products:
P(t) = g (A(t), O(t), M(t), D(t))
where P is polarisation, A is algorithmic amplification, O is outrage, M is misinformation and D is existing divisions at time t. Other things being equal, more amplification, outrage, misinformation or division means more polarisation. The outrage that raises engagement is the same outrage that, once amplified, raises polarisation. What increases the platform's revenue may also increase society's cost. However, this is an externality. The platform captures private return, while society bears costs in distrust, communal tension, reputational destruction and weakened public discussion. Because the platform does not pay these costs, it has no commercial reason to count them when deciding what to rank. A factory that profits while polluting neighbouring communities is the familiar parallel.
There is no denying that the analogy has its limits. Digital communication is more complex than pollution; its social costs are harder to measure, and social media generates substantial benefits: cheaper communication, access to customers for small businesses, wider access to information, and a platform for voices often excluded by traditional media. The question, therefore, is not whether social media is good or bad, but whether market incentives adequately account for the wider consequences of how attention is generated. When they do not, the market will tend to produce more socially costly engagement than society would otherwise choose. This is also why regulating individual speech alone cannot solve the problem: removing one harmful item leaves intact the mechanism that determines what becomes visible, amplified and rewarded next.
Polarisation as Both Outcome and Input
The relationship is dynamic. Divisions shape what people watch and share; recommendation systems learn what draws strong responses and supply more of it; exposure strengthens attitudes, which shape the next round of clicking. Polarisation in one period becomes an input in the next:
A(t+1) = h (P(t), E(t), S(t))
D(t+1) = k (D(t), P(t))
where E is engagement, S is the recommendation system's design (in particular, the objective it maximises), and t+1 is the following period; amplification rises with both polarisation and engagement. The second equation closes the loop: divisions are not fixed background but are reinforced by the polarisation they help produce.
I call this a matrix because the variables do not form a chain from cause to effect. Each affects several others and is affected in turn: users influence algorithms, algorithms influence users, and engagement reinforces divisions. The claim that social media simply causes polarisation misdescribes this structure, and so does the claim that it plays no part. This is important and matters for Bangladesh.
Nahil's figures, about 64 million social-media identities, roughly 36 per cent of the population, describe a digital space that is a newspaper, marketplace and neighbourhood meeting place at once, operating amid intense political competition, religious sensitivity and dense settlement. There, speed of circulation can matter as much as possible. The same logic may apply to moderation, though this is a hypothesis that data could test: where advertising revenue per user is modest, the platform's private return from reviewing Bangla-language content is likely to be smaller, while the social cost of neglect is not correspondingly smaller. A correction tends to be less engaging than the rumour it corrects, so the loop can turn several times before the truth arrives.
Changing the Incentive Structure
If the problem is partly economic, policy should address incentives as well as content through two kinds of measures.
The first enables measurement: transparency requirements, privacy-protected researcher access to platform data, and Bangladesh-specific reporting on Bangla-language content. A cost that cannot be measured is hard to debate or regulate sensibly. Bangladesh needs evidence on how such content circulates, what receives amplification, and how quickly falsehoods spread.
The second changes the payoff. A genuine choice between algorithmic and chronological feeds gives users an exit from the engagement-maximising default, and friction before rapid sharing raises the cost of spreading questionable material when it spreads fastest. The most direct instruments make platforms answer for their own systems: duties to assess and reduce the systemic risks of recommendation design, backed by penalties, as in the European Union's Digital Services Act. Because such costs are hard to quantify, a calibrated charge is unrealistic; these duties are a practical second-best. To protect free expression, responsibility should attach to platform design and amplification decisions, not users' lawful opinions.
The objective is not to eliminate disagreement, on which democracies depend, nor to give governments power over legitimate opinion. It is to narrow the gap between the private rewards of engagement and the social costs certain engagement imposes.
Because attention is sold, businesses compete to capture it. The difficulty begins when the most profitable way of capturing attention is not the most socially beneficial way of informing citizens. Platforms capture much of the return from engagement while society absorbs part of the cost of polarisation. Until that divergence is addressed, regulating only what individuals say will keep treating the symptoms while leaving the structure untouched.
The writer is a former physicist and nuclear engineer at Bangladesh Atomic Energy Commission and a Professor Emeritus of Economics at Eastern Michigan University, USA. He can be reached at [email protected]