The Trap of Immediate Metrics
Here's something most of us don't notice until it's too late: we've trained ourselves to value what moves fast. Quarterly returns. Productivity dashboards. Click-through rates. The things that show up in bold on a spreadsheet get our attention first, and over time, they become the only things we see.
Cornelia Walther puts it bluntly in her Psychology Today piece — trust and dignity aren't optional extras bolted onto the "real" economy. They're load-bearing walls. But try telling that to a boardroom where the quarterly report is due in three hours and nobody's talking about whether the team has enough oxygen to keep showing up.
The problem isn't that metrics are useless. They coordinate complex systems, sure. The problem is what happens when they stop being tools and start being masters. We begin to treasure what we can easily measure instead of measuring what we actually treasure. It's a quiet inversion, and it's been happening for decades.
Take care work. Raising children. Supporting elders. Listening someone through a breakdown. Mentoring a junior colleague who's clearly drowning. Preventing harm before it happens. None of this shows up on a balance sheet, even though everything else depends on it. The International Labour Organization estimated that 708 million women worldwide were outside the labor force in 2023 because of unpaid care responsibilities. The World Inequality Report 2026 shows women work longer hours than men once you count domestic and care work — and receive roughly a third of total labor income.
We built systems that depend on care while teaching ourselves to undervalue it. That's not a bug. It's the design.
And now AI is walking into that landscape like a mirror — reflecting exactly what we've already decided matters, then amplifying it with terrifying efficiency.
AI as an Intention Accelerator
Let's be honest about what AI actually is. It doesn't arrive with discernment. It isn't born compassionate. It learns human patterns, optimizes whatever goals we give it, predicts behavior, and scales instructions with a speed that makes most of us uncomfortable.
A company that treats people mainly as cost centers will use AI to monitor, rank, and replace. A school system obsessed with standardized performance will use AI to narrow learning down to what fits on a scantron. The danger here isn't biased data or faulty algorithms — those are real problems, but they're symptoms. The deeper danger is impoverished purpose.
Walther calls AI a "mirror and multiplier." I think that's right, but it undersells how fast the amplification happens. Once you embed a narrow definition of value into code, datasets, and dashboards, AI doesn't just follow that path — it carves a highway. And before you know it, the organization has no idea how to think about anything else.
This is where prosocial AI becomes more than a buzzword. It asks builders and users to examine their personal values before those values get tacitly embedded in code. Before datasets are selected. Before dashboards are designed. Before the business model locks into a prison of value fixation that forgoes everything that actually matters.
The UNDP Human Development Report 2025 frames this perfectly. Subtitled "A matter of choice: People and possibilities in the age of AI," it argues that development in this era depends more on mobilizing human choices and expanding freedoms than on the technical capabilities of models themselves. The future of AI, in other words, is a question of human agency.
A society that measures mainly short-term material indicators will build AI for speed, scale, prediction, persuasion, substitution, and control. A society that values human flourishing will build AI for agency, dignity, learning, care, social connection, prevention, and planetary responsibility.
Both are technically possible. The difference is what we decide to notice first.
The Toll of Disconnected Systems
Here's a number that should keep you up at night: loneliness accounts for approximately 871,000 deaths each year. That's roughly 100 deaths an hour. The WHO Commission on Social Connection, established in November 2023 and co-chaired by Dr. Vivek Murthy, published this finding in their landmark 2025 report.
One in six people worldwide experience loneliness. Among adolescents and young adults, it's closer to one in five. In lower-income countries, nearly one in four.
We talk about AI and productivity like they're the same conversation. They're not. The real story is what's happening to human connection while we're all optimizing for efficiency. High productivity curves coexisting with exhausted workers. Booming platforms deepening anxiety. Profitable products weakening attention. Smart systems reducing friction while increasing dependence.
The WHO report doesn't just document the damage — it maps solutions. Advocacy campaigns. National policies. Community strategies. Individual relationship work. Eight WHO Member States have already adopted policies on social connection, which is promising but still a fraction of what's needed.
What's striking is how this connects to everything else in the article. Social isolation isn't just an individual mental health issue. It has negative impacts on education, employment, economic growth, and innovation. The economic costs to employers, healthcare systems, and individuals are significant — and only beginning to be estimated.
When we design AI systems that optimize for engagement metrics, we're often optimizing for exactly the kind of isolation the WHO is warning about. Algorithms that keep you scrolling don't build community. They simulate connection while quietly eroding the conditions that make real connection possible.
The question isn't whether technology magnifies intention. It does. The question is what intention we're feeding it.
The A-Frame Mindset Checklist
Walther offers a practical framework for this work — the A-Frame mindset. Four entry points. Not a technical spec. A way of paying attention.
Awareness. See the values already shaping your choices. Ask what your metrics reward, what they ignore, and how those invisible incentives influence behavior. Most of us don't realize we're operating on autopilot until something breaks.
Appreciation. Recognize forms of value that are easy to overlook. Care is intelligence. Relationships are infrastructure. Attention is finite. Human agency must be cultivated, not assumed.
Acceptance. Face the limits of short-term indicators. Numbers can guide decisions, sure. But they cannot carry the full meaning of a life, a community, or a planet. AI can process patterns; it cannot replace conscience, responsibility, or lived wisdom.
Accountability. Align tools with values. Ask who benefits, who carries the burden, who can challenge the outcome, and what long-term effects are being created.
The next frontier isn't inside a server farm. It's situated inside us. We need to retrain our individual and collective attention, measure what matters, and protect what must first be noticed before it can be valued and amplified.
Care itself is a form of intelligence — not just emotional generosity. To care well requires perception, timing, restraint, memory, empathy, judgment, and imagination. A caregiver reads incomplete signals. A teacher senses when a student is present in body and absent in spirit. A nurse notices the difference between a symptom and fear. A good manager recognizes when high performance masks exhaustion.
This intelligence is relational, embodied, and contextual. It cannot be fully captured in a metric. But it can be supported — or damaged — by the systems around it.
Prosocial AI can serve humans imbued with this kind of caring intelligence. It can reduce administrative load, detect patterns early, widen access to knowledge, translate complexity, and help people make wiser choices. It can help institutions see where stress accumulates, where exclusion repeats, where support is missing, and where prevention would be wiser than repair.
Yet AI becomes prosocial only when guided by a mature human mindset. The same dashboard that reveals burnout can be used to blame workers. The same prediction model that identifies need can be used to deny services. The same personalization engine that supports learning can be used to manipulate desire.
Technology magnifies intention. That's why mindset is infrastructure.