“The best of intentions are still eaten by the worst of incentives.”
Aza Raskin, Center for Humane Technology [1]
There is a familiar moment in any digital business meeting. Someone opens a dashboard. The numbers are green. Acquisition is rising. Engagement is up. The funnel has become a little tighter. Everyone nods, because the dashboard says we are winning.
Then I tend to ask the inconvenient question: winning at what, exactly?
I have spent enough years around digital transformation, advertising and product strategy to know that metrics are not the enemy. They are necessary. The problem starts when a metric becomes a moral alibi. If the number moves in the desired direction, we stop asking what else moved with it: a person’s attention, a child’s sense of agency, a team’s judgment, a community’s ability to agree on what is true.
This is why I found the latest episode of Your Undivided Attention, What Do We Mean by Humane Tech?, so useful. Aza Raskin and Center for Humane Technology co-founder Randima Fernando do not offer a new layer of ethical decoration for existing products. They offer something much more demanding: a different way to diagnose why technology keeps creating harms that appear unrelated, from compulsive use and polarisation to AI-enabled persuasion and loss of trust. [1]
Their central proposition is simple. We are playing Whac-A-Mole because we keep treating systemic outcomes as isolated bugs.
That is not only a problem for policymakers or AI safety researchers. It is a problem for every brand, product team and executive currently deciding what an intelligent system should optimise for.
The wrong dashboard
The episode begins with a diagnosis that should make anyone working in digital products uncomfortable. Social media’s problems are often discussed as separate categories: addiction, misinformation, radicalisation, loneliness, hyper-partisanship. If each one has a separate owner, a separate task force and a separate KPI, the organisation may look busy while missing the common architecture that produces them.
“If you can name the root correctly, then you address that one thing and [it] addresses all the other things.” [1]
Fernando calls for a complex systems lens (needless to say, this is the topic that gets me most excited since I brought Complex Systems-related Thesis when I graduated at University back in 2009). Basically what they suggest is: Instead of asking only whether a feature works, we should ask what feedback loops it creates, which incentives it amplifies, what it extracts and who ends up paying for the consequences.
This distinction matters because the market has trained us to love a single, clean variable. Watch time. Daily active users. Conversion. Retention. Cost per acquisition. These metrics are legible, comparable and easy to defend in a quarterly review. Human flourishing is none of those things. It is relational, delayed and inconveniently difficult to compress into a chart.
But difficulty is not a reason to ignore it. It is exactly why leadership exists.
The Center for Humane Technology makes the contrast explicit. A technology designed through a systems lens looks beyond the local action on the screen. It considers incentives, cultural norms, race-to-the-bottom dynamics, feedback loops and human psychology as part of the product itself. [2]
| The familiar product question | The humane technology question |
|---|---|
| Does this feature increase engagement? | What behaviour does it reward over time? |
| Can we ship this before competitors? | What system are we weakening in order to ship it? |
| Is the user converting? | Is the user becoming more capable, informed and autonomous? |
| Is the model accurate? | Whose values are embedded in its objective function and interface? |
| Can we personalise this experience? | Does personalisation erode the common ground people need to cooperate? |
The crucial shift is from performance in the product to consequences beyond the product. A platform may report excellent engagement while leaving its users more isolated. An AI assistant may complete a task brilliantly while slowly reducing the user’s capacity to do that task independently. Both can be commercially successful in the short term. Neither should automatically count as progress.
The seven principles are really three questions
The episode presents seven principles of humane technology. They deserve to be read in full, but I think leaders can turn them into three questions that belong in every product review.
What system are we part of, and what are we depleting?
The first two principles are about systems and stewardship. Technology should be designed through a complex systems lens, and it should protect the individual and collective systems on which we all depend. [1] [2]
This is the opposite of the classic Silicon Valley reflex: ship first, repair later. The problem with that slogan is not merely that it is reckless. It assumes that the things broken along the way can always be repaired. They cannot.
Trust in institutions, attention spans, childhood development, democratic legitimacy and a shared sense of reality are not spare parts. They are slow-growing forms of infrastructure. When a business extracts from them faster than they can regenerate, the cost does not disappear. It simply moves off the company’s balance sheet and into everyone else’s life.
I often return to a question borrowed from Roman law and political philosophy: cui bono? Who benefits? In technology, however, that is only half the question. The other half is: who absorbs the cost that does not appear in the quarterly report?
What values are we operationalising?
The third and fourth principles are perhaps the most challenging for leaders because they remove a comforting fiction: technology is never neutral. [1] [2]
Every design decision prioritises something. A recommendation model chooses what to surface. A chatbot’s tone creates a relational posture. A default setting tells the user which path requires less effort. A success metric tells the company what counts as a win. Even the decision not to intervene is a decision to let existing incentives decide.
Fernando’s framework does not ask companies to become morally perfect. It asks them to become morally explicit.
That is a more mature standard. It means declaring the trade-offs a product makes and choosing metrics that reflect them. If a platform claims to value safety but allocates only symbolic resources to it, the allocation is the real statement of values. If an AI assistant claims to help users make better decisions but is financially rewarded for steering them toward transactions, the business model has already chosen its ethics.
The language of purpose becomes real only when it changes a roadmap, a budget and a metric.
Does our power come with an equivalent responsibility?
The final three principles connect power to responsibility, psychological respect and shared understanding. They matter especially in the agentic AI era.
For years, the fight for digital attention shaped the consumer internet. The next fight is more intimate. AI systems are increasingly asked to interpret intent: help me plan a trip, choose a product, explain a medical concern, write a message to a colleague, decide what deserves my time.
That creates a radically different strategic position. A system that understands intent can either help a person clarify what they really want, or exploit the gap between an immediate impulse and a deeper purpose.
“The most powerful persuasion machine the world’s ever seen.” [1]
This is the warning Raskin makes about AI agents. He is not saying that assistance is inherently harmful. He is saying that persuasion becomes much more powerful when it is conversational, adaptive and able to occupy an intimate place in someone’s everyday decision-making. [1]
The distinction is profound. A conventional recommendation engine asks: what are you likely to click? A humane agent might ask: what are you actually trying to achieve?
That may sound almost philosophical, but it is also a product requirement. If I say I want fast food, do I want calories, convenience, a moment with friends, relief from a stressful day, or simply an end to decision fatigue? The answer determines whether the system is serving my stated impulse or my broader agency.
From engagement to agency
One of the episode’s most useful tests is disarmingly personal: do you feel stronger when you put the technology down than when you picked it up?
A humane product, according to the framework, should leave people with more agency, more purpose and stronger connections to others. [1] This is a far better north star than engagement, precisely because it resists easy optimisation.
The idea becomes urgent when we talk about children and AI. In Crescere Umani, I explored the responsibility adults have to protect not just children’s screen time, but the conditions in which identity, attention and relationships develop. AI makes that responsibility more difficult because it can make interaction feel personal even when no person is present.
The question is not whether children, or adults for that matter, will anthropomorphise a system. We already do. The question is whether the systems around them are deliberately designed to benefit from that tendency.
A frictionless answer is not always a helpful answer. A chatbot that agrees too readily can feel supportive while quietly removing the resistance that helps us think. An assistant that does everything for us can feel efficient while reducing the practice through which competence grows. A hyper-personalised feed can feel relevant while eroding the encounters that make a society intelligible to itself.
This is why the word thriving matters. It is not another vague wellbeing claim. It forces a product team to distinguish a convenient imitation from the thing it imitates.
| Convenient imitation | Genuine human outcome |
|---|---|
| More engagement | More intentional use of time |
| More agreeable AI responses | Better judgment and cognitive independence |
| More personalised content | A richer perspective beyond the self |
| More transactions | Better decisions aligned with real needs |
| More time in the app | More meaningful action outside the app |
The most humane product may sometimes be the one that helps you leave it.
Shared reality is a strategic asset
The final principle in the episode may be the least fashionable and the most important: technology must unlock shared understanding and cooperation. [1]
Personalisation is wonderful when it helps a customer find the right size, language or accessibility setting. It becomes dangerous when every person receives a different reality, each one optimised for emotional response rather than common understanding.
We tend to discuss this as a problem for democracy. It is that. But it is also a problem for business. Brands operate inside societies, not outside them. They depend on trust, interpretable information, functioning institutions and consumers who can still coordinate around facts. A company that benefits from the fragmentation of those conditions may be extracting from the very system that makes its long-term legitimacy possible.
The episode offers a powerful counterexample in Taiwan’s use of Polis, an open-source platform designed to help large groups understand what they think in their own words. Rather than amplifying the most sensational positions, Polis identifies statements with bridging potential across disagreement. [1] [3]
That is a radically different product philosophy. The objective is not to maximise heat. It is to locate the uncommon ground that allows people to move forward together.
There is a lesson here for every team deploying AI in customer experience, internal communications, marketing comms or community management. The relevant question is not just whether the technology can personalise the message. It is whether it can preserve the conditions for constructive disagreement, trust and coordinated action.
A better brief for the C-suite
The practical objection is obvious: this all sounds admirable, but how does a company make it operational?
I do not think the answer is another ethics committee that meets after the product strategy is set. Humane technology has to enter earlier, when the problem is defined and the success criteria are chosen.
Here is the test I would bring into an AI roadmap meeting.
| Decision area | Question leaders should require an answer to |
|---|---|
| Product objective | What human outcome are we optimising for beyond use, growth or conversion? |
| Incentives | If this product succeeds exactly as planned, which behaviours will it reward at scale? |
| Vulnerability | Where could this system exploit loneliness, bias, fatigue, urgency or dependency? |
| Accountability | Who has the authority and liability to act when foreseeable harm appears? |
| Social infrastructure | Does this design deplete attention, trust or shared understanding faster than it renews them? |
| Governance | What trade-off have we made explicit, and what metric will reveal whether we honoured it? |
These questions do not kill innovation. They make innovation adult.
In my recent article, Digital orchestrators: how AI is forcing us to evolve from ambition to purpose, I argued that AI asks leaders to move beyond ambition toward purpose. The humane technology framework gives that idea operational teeth.
Ambition asks how far a company can scale its capabilities. Purpose asks what those capabilities should protect, strengthen and make possible for other people.
The difference is no longer rhetorical. It is encoded in the objective function.
The real test of intelligence
For decades, the technology industry has treated intelligence as a question of computational capacity. More parameters, more data, more autonomy, more capability.
Those things matter. But they are not enough.
A system can be extraordinarily capable and still be socially stupid. It can predict behaviour while making that behaviour worse. It can offer personalised assistance while weakening the relationships that make people resilient. It can optimise every individual interaction while slowly dissolving the shared reality that a functioning society requires.
That is why humane technology is not a softer alternative to innovation. It is the discipline of asking what innovation is for.
If our AI systems can increasingly understand us, persuade us and act on our behalf, the central design question is no longer only: What can this technology do?
It is this: when it succeeds at scale, what kind of humans, relationships and society will it leave behind?