Why the Poorest Countries Are the Most Hopeful About AI

Ask a software engineer in San Francisco how they feel about artificial intelligence, and you may get anxiety: job losses, deepfakes, a machine that writes better than they do. Ask a small-business owner in Lagos or a student in Jakarta, and you are far more likely to hear hope. This is not a small gap. Across global surveys, the people most optimistic about AI live in poorer, faster-growing countries, while the richest, most technologically advanced societies are the most worried. The pattern is consistent enough that it needs explaining.

The model: reference dependence

Humans judge their situation relative to a reference point, not in absolute terms. This is reference dependence, the core of prospect theory, developed by psychologists Daniel Kahneman and Amos Tversky in 1979 — work that won Kahneman the 2002 Nobel in economics and became one of the most replicated findings in behavioural science. Their experiments showed that the same outcome feels like a gain or a loss depending entirely on the baseline you measure it from. A paired finding, loss aversion, adds that losses hurt roughly twice as much as equivalent gains feel good.

Apply that to AI and the paradox dissolves. A new technology is not judged on what it is. It is judged against the baseline it lands on, and against what it threatens to take away.

• Where the baseline is good — reliable doctors, functioning banks, decent schools, stable jobs — a powerful new tool mostly threatens to disrupt things that already work. The downside is vivid; the upside is marginal.
• Where the baseline is poor — no specialist within a hundred miles, no legal advice you can afford, patchy schooling — the same tool offers things you simply could not get before. The upside is enormous; there is less to lose.

Optimism about AI, in other words, tracks how much the status quo was failing you.

The mechanism, step by step

1. The marginal gain is larger where provision is thinner. An AI tutor adds little to a child with excellent teachers. It transforms the prospects of a child with none. The same chatbot is a toy in one place and a lifeline in another.
2. The threat is smaller where there is less to protect — and loss aversion makes that threat loom large where it exists. AI anxiety in rich countries centres on white-collar jobs, professional status, and trusted institutions. Loss aversion means the prospect of losing those weighs heavier than any gain. In economies where formal professional employment is scarce, there is less incumbent comfort for the technology to erode.
3. Expectations are shaped by recent direction of travel. People in fast-growing economies have watched mobile phones, mobile money, and cheap internet leapfrog infrastructure that the West spent a century building. They have a lived template for technology arriving fast and improving life. That template resets the reference point upward: they expect the same from AI.

Put together: the groups with the most to gain and the least to lose, who also have recent memory of technology delivering, are the enthusiasts. The groups with functioning systems to defend and good reason to fear disruption are the sceptics.

The other view

There are honest competing explanations, and they are not mutually exclusive.

One is exposure and information. Rich-country publics have had longer, deeper contact with AI's failures — hallucinating chatbots, algorithmic bias, misinformation — and their worry may reflect better-informed judgement, not a worse baseline. On this reading, optimism elsewhere is partly naïveté that experience will erode.

A second is media and political framing. Western coverage of AI is dominated by job loss, regulation, and existential risk. Where that framing dominates public discussion, fear is partly a product of the stories people hear, not a personal cost-benefit calculation.

A third is who gets surveyed and what "AI" means to them. A respondent imagining a helpful phone assistant and one imagining mass redundancy are answering different questions. Some of the gap may be measurement, not sentiment.

What is fact: multiple cross-country surveys show markedly higher AI optimism in lower- and middle-income countries than in rich ones. What is interpretation: why. Reference dependence is the cleanest single explanation, but exposure, framing, and measurement all plausibly contribute. What is unknown: whether the optimism survives contact with AI's real-world downsides as they accumulate.

Where else this pattern shows up

The same logic governs reactions to almost any disruptive change.

• Mobile banking. M-Pesa in Kenya was embraced precisely because the alternative — formal banking — was absent for most people. In countries with ubiquitous branches and credit cards, the same innovation felt redundant.
• Telemedicine. A breakthrough where the nearest doctor is a day's travel; a downgrade where in-person care is the norm.
• Your own career. A junior employee welcomes a reorganisation that a senior one dreads — the junior has position to gain and little to protect, while the senior has a comfortable role to lose. Enthusiasm for change is rarely about the change itself. It is about where you are standing when it arrives.

The rule worth carrying away: attitudes to a new thing are a map of people's baselines, not a verdict on the thing. When you see sharply divided opinion about an innovation, ask first what each side is measuring it against.

What to watch

The decisive signal is what happens to developing-world optimism as AI's downsides arrive there too — scams, misinformation, displaced informal work, dependence on foreign-owned systems. If reference dependence is the main driver, enthusiasm should hold as long as the practical upside keeps outrunning the harm. If exposure is the main driver, optimism should fade toward Western levels as contact deepens.

Expectation: ~65% that AI optimism in lower- and middle-income countries remains significantly higher than in rich countries through the next several years of survey data — because the gap in baseline provision, the real source of the divide, will not close nearly as fast as the technology spreads.

Distilled from The Economist — International