The shift to ensure that we use AI without losing ourselves. #
Posted July 27, 2026 [ Reviewed by Michelle Quirk
](/us/docs/editorial-process)
Key points
- Greater confidence in generative AI has been associated with less critical-thinking effort.
- "Return on values" widens the picture across purpose, people, prosperity and planet.
- We must align our actions with the future we want, then align our algorithms with that direction.
Most humans no longer experience artificial intelligence (AI) as dramatic. What felt like a grand technological revolution in 2022 is gradually becoming a standard asset in our life, labor, and leisure. We ask a chatbot to improve an email, summarize a report, suggest what to buy, or help us prepare for a difficult conversation. At work, an AI tool may rank applicants, score performance, or recommend which customer deserves attention.
Together, these choices begin to shape how we think, decide, and relate to one another. The central question is simple: Does AI strengthen our ability to act well, or gradually make important choices for us?
AI learns from the world we have built #
AI systems learn from human language, past decisions, and institutional data. They also absorb the priorities behind those records.
Imagine a customer-service department that says it values care, yet rewards staff mainly for ending calls quickly. An AI assistant trained around that system will probably learn to shorten conversations. It may improve efficiency while leaving customers feeling unheard.
A recruitment tool can repeat old preferences. A learning platform may help students reach answers without ensuring understanding. A recommendation system can discover that outrage keeps people engaged.
The system may be functioning exactly as requested. The deeper issue lies in what people and institutions choose to reward.
This is where ProSocial AI begins: before coding or deployment. It asks whether our stated values are visible in our behavior, incentives, and decisions.
Convenience changes behavior #
Human beings naturally conserve effort. AI can free attention for other tasks. The outcome depends on what we hand over.
A 2025 Microsoft Research study of 319 knowledge workers found that greater confidence in generative AI was associated with less critical-thinking effort. It also found that AI moved thinking toward checking information, integrating responses, and supervising the task. AI changes where human judgment is needed. Users require the skill and confidence to occupy that space.
A 2026 study in Scientific Reports offers a related warning. Participants judged whether faces were real or AI-generated while receiving guidance that was correct only half the time. People with more positive attitudes toward AI were more strongly influenced by its advice and became less effective at distinguishing real from synthetic faces. Polished output can make weak advice feel authoritative.
Human oversight therefore requires more than a final approval button. The person must understand the task, notice an error, and feel able to challenge the result. A teacher who cannot explain an automated score cannot defend a student. A manager who no longer understands the process cannot judge whether the outcome makes sense.
Double alignment: get the order right #
A healthy relationship with AI requires two alignments.
The first is between aspiration and action. We clarify what we are trying to achieve, then examine whether our behavior supports it.
A school that values independent thought should reward curiosity and reasoning. A company that values well-being should reflect it in workloads and promotion decisions. A public agency that promises fairness should make decisions understandable and open to appeal.
The second alignment is between those clarified aspirations and the algorithm. Once the human purpose is clear and visible in practice, AI can be designed to support it.
When an organization buys an AI system first, people often reshape their goals around what the tool can measure. Teachers teach to the dashboard, employees work for the score, and managers confuse what is countable with what is valuable.
IntelligenceEssential Reads
Four questions before using AI #
A person, team, or institution can ask four questions:
Pursuit: What human outcome are we trying to improve?Power: Which part of this task (dis)empowers people?Proof: How will we know whether the system is helping or harming humans and/orthe environmentover time?Prospect: Who benefits, who carries the risk, and what resources does the system consume?
These questions change everyday use. Someone writing a difficult message might first decide what they genuinely want to communicate, then use AI to test tone. A student might attempt a problem before asking AI to identify gaps. A manager could use AI to spot patterns while keeping promotion decisions grounded in evidence and conversation.
Research suggests that structure matters. A 2025 Scientific Reports study involving preservice teachers found positive links between AI use and academic achievement when accompanied by genuinely shared reflection and managed cognitive off. Another
2025 experimental studyfound that AI collaborationimproved immediate task performance while sometimes reducing
intrinsic motivationand increasing boredomin later solo work. These findings support selective use that protects learning and human engagement.
From return on investment to hybrid return on values #
Organizations usually assess AI through return on investment: time saved, costs reduced, and revenue generated. These measures matter. They leave major effects invisible. It is time to shift from treasuring what we can easily measure, toward measuring what we must treasure to build a hybrid future that is pro-people, pro-planet, and pro-potential.
An AI system may save staff hours while weakening expertise, increase sales while encouraging manipulation, or speed recruitment while making applicants feel powerless.
Return on values widens the picture across four areas: purpose, people, prosperity, and planet.
Purpose asks whether the system addresses a worthwhile need.People covers agency, dignity, capability, trust, and inclusion.Prosperity examines durable value and how benefits and risks are shared.Planet considers energy, water, hardware, and waste.
The [International Energy Agency’s 2025 Energy and AI report](https://www.iea.org/reports/energy-and-ai/) and its
[2026 update](https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary)show why the physical footprint of AI belongs in ordinary investment decisions.
This means making important effects visible before a decision and monitoring them after deployment. Consider a company introducing an AI assistant. A conventional assessment might record that employees complete reports 20 percent faster. A return-on-values assessment would also ask whether the reports are better, whether staff retain the relevant expertise, whether junior employees still learn, whether errors create work for colleagues, and whether the system deserves the energy and resources it consumes.
The same logic applies at home. An AI tool may save time when planning a trip or composing a routine message. Using it to settle a disagreement, interpret another person’s motives, or make a major life decision carries a different weight. The more closely a task touches identity, relationships, or responsibility, the more human attention it deserves.
The human choice inside every prompt #
ProSocial AI is built through many small choices. Think before prompting. Use AI to widen options while keeping the goal human-led. Check high-stakes claims. Preserve a human route for challenge and appeal. Notice when a skill is weakening through disuse. Ask whether convenience is becoming dependence.
The aim is hybrid intelligence: human values, ethical judgment, and artificial capability strengthening each other. AI brings speed, scale, and pattern recognition. People bring aspiration, context, conscience, and responsibility.
The order remains decisive: Align our actions with the future we want, then align our algorithms with that direction. AI will learn from us. We should give it better patterns to learn from.