Tech for Social Good: Why Technology Is Not Progress by Default

Nobody asked for supermarket self-checkout counters. They replace cashiers, they are rarely better for the customer, and the productivity gain is marginal at best. That gap, between shipping technology and producing progress, is the whole problem with tech for social good. As a product manager at a Silicon Valley-backed scale-up, I watched daily user numbers in the millions and felt how abstract decisions with that reach become. Which of the two you get, technology or progress, is a choice someone makes in a product meeting, not a law of nature. This essay makes the case for that with economic evidence, then lays out the three-step test I use to judge whether a product is worth building.

What Counts as Tech for Social Good

Tech for social good is technology built so that its benefits flow to society broadly: to the people who use it, the people who work with it, and the people affected by it, not only to the company shipping it. The definition has a trap in it, though. "Tech for good" describes an intent. Whether a product does good is an outcome, and the outcome depends on design choices: what gets automated, who gets more productive, and who captures the gains. A hackathon label or a CSR page says nothing about any of those.

The Comfortable Assumption: Technology Drives Progress

Start with the strongest version of the optimist case, because it is genuinely strong. Global poverty has fallen dramatically over recent decades. Renewable energy became cheaper than fossil fuels. Access to electricity, child mortality, life expectancy: the trend lines run the right way, and Hans Rosling's Factfulness documents how consistently most people underestimate them. Technology sits behind every one of those curves.

So does technology drive progress? Sometimes, enormously. The mistake is reading "sometimes" as "inevitably". The same decades that produced those curves also produced technologies that made almost no one better off, and the difference between the two kinds is measurable.

So-So Technologies: When Innovation Helps No One Much

Nobel laureates Daron Acemoglu and Simon Johnson call them "so-so technologies" in Power and Progress: innovations that displace workers without a productivity gain large enough to create new work or better service. The supermarket self-checkout is their textbook case, and it is worth walking through, because it looks like progress from the P&L and like nothing from everywhere else. The cashier's job disappears. The customer now does the scanning, unpaid. The queue is not shorter. The productivity gain is a rounding error, and it flows entirely to the retailer's labour line.

Compare that with the railway. It destroyed the work of coach drivers, and it enabled entire new industries, logistics and tourism among them, which created far more work than it destroyed. Same pattern, opposite outcome:

Who captures the gains is not a side question. It is the question:

"Automation has also been a major booster of inequality because it concentrates on tasks typically performed by low- and middle-skill workers in factories and offices. Almost all the demographic groups that experienced real wage declines since 1980 are those that once specialized in tasks that have since been automated."

Acemoglu & Johnson, Power and Progress

The empirical backing is hard to argue with: Acemoglu and Restrepo's Econometrica study attributes 50 to 70% of the changes in the US wage structure over four decades to the relative wage declines of workers specialized in routine, automatable tasks. Meanwhile the World Inequality Report finds the top 1% share of US wealth approaching 35% in 2020, near its Gilded Age level. Anne Case and Angus Deaton document where the other end of that distribution went in Deaths of Despair: as manufacturing work disappeared, deaths from suicide, overdose and alcoholism rose sharply among Americans without a college degree.

If AI Pays for Everything, What Do People Do?

The standard tech-optimist answer to all of this is universal basic income: let AI create so much value that everyone gets paid to do whatever they like. I used to find that tempting and I no longer do, for a reason that has nothing to do with the budget math. A job is not only an income. Martin Seligman's well-being research at Penn's Positive Psychology Center puts accomplishment among the five pillars of human flourishing: people pursue achievement and mastery for its own sake, even when it brings no other reward. A transfer payment replaces the salary and none of the rest: the tasks, the competence, the being needed.

That is why the first question about any labour-displacing technology is not "how do we compensate the displaced" but "what will they do". Inside companies, the same question has a constructive answer, and I wrote it up separately: upskilling employees so the technology raises their output instead of replacing it.

A 3-Step Framework for Building Tech for Social Good

The test I run before building anything, assembled from the economics above and from years of product work:

1. Build for humans first, and check who captures the gains

Increase the productivity of tasks humans are already performing, rather than deleting the human from the task.

"When a skilled artisan is given a better chisel or an architect has access to computer-aided design software, their productivity can increase significantly."

Acemoglu & Johnson, Power and Progress

Then ask the uncomfortable follow-up: if this works, where does the gain go? To the user's output, the worker's wage, or only the P&L? If the honest answer is the third, you are building a so-so technology with a mission statement attached.

2. Create new tasks, not just savings

The railway test. A technology earns "net-positive" when it opens work and capability that did not exist before, for skilled and unskilled people, not when it books the same output at lower headcount. For anything AI-shaped, that means designing the new tasks and the training into the product plan, not leaving them as an externality for society to absorb.

3. Widen human decision-making

The third pattern that reliably produces good outcomes: technology that gives people access to information and judgement they could not reach alone. Decision-making is constrained by accurate information more than by intelligence, and tools that fix the information constraint make everyone who touches them more capable. That is the standard I hold my own products to, and the reason I keep writing about understanding how the world works as the precondition for having any impact on it.

What It Takes: A Tech Sector That Questions Itself

I believe a reflective view on tech within the tech sector is the first necessary step, and it is the piece I still see missing. Working in tech for years, from Europe to the US, I heard almost no one seriously questioning the impact of their own product decisions, mine included. The incentive structure explains it: the main mandate of a big tech company is maximizing shareholder value, and the impact question has no owner in the org chart.

I have taught AI for business development to executives and board members across Europe, and published research on generative AI for growth, and that work keeps returning me to the same conviction: the people building and deploying this wave have more choice about its shape than the "technology is destiny" framing admits. Entrepreneurship is the counterweight I bet on. New companies can use AI to challenge incumbents instead of entrenching them, and movements like micro SaaS and indie hackers show a generation of builders who do not need investor-scale returns to justify building with intention.

Technology is not progress. It is raw material for progress, and the difference is decided by people who choose what to build and how. The books that shaped this argument, alongside everything else I read, sit on my inspiration page.

---

Sources

• Daron Acemoglu & Simon Johnson, Power and Progress
• Acemoglu & Restrepo, Tasks, Automation, and the Rise in US Wage Inequality, Econometrica
• World Inequality Report 2022, chapter 4
• Case & Deaton, Deaths of Despair and the Future of Capitalism
• Our World in Data on poverty and renewables
• Penn Positive Psychology Center on PERMA