In our rush to embrace AI and data-driven decision making, we're making a fundamental error that could have profound consequences for how we think, work, and live together. We're building systems that prioritize "human + machine" interactions when what we truly need is "human + human + machine" frameworks.

The Difference Matters

The distinction is subtle but crucial. Throughout human history, our greatest achievements and resilience have come through collective intelligence—people thinking together, challenging each other, providing emotional support, and creating shared meaning. These social processes aren't just nice-to-have features; they're the foundation of what makes us human.

The "human + machine" paradigm isolates individuals with technology, while the "human + human + machine" approach preserves the social fabric that has been essential to human flourishing.

False Conviction and Distributed Harm: The Self-Driving Car Paradox

What makes our current AI trajectory particularly concerning is the false conviction people develop when interacting with AI systems. We've established rigorous safety standards and public scrutiny for self-driving cars because the risk is obvious: one malfunction could cause immediate, visible harm.

Yet we're not applying the same scrutiny to AI systems that influence our information, decisions, and values.

Unlike the dramatic crash of a self-driving vehicle, AI systems like large language models operate through what I call "death by a thousand paper cuts" - a fundamentally different harm model that's:

• Distributed across millions of daily interactions
• Often subtle and impossible to trace back to a single source
• Cumulative in their societal impact over time
• Targeting our information ecosystem, decision-making processes, and social bonds

Consider how we interact with these systems: A physician relies on an AI diagnostic tool without consulting colleagues. A judge reviews an algorithm's sentencing recommendation without community input. A student crafts essays with AI assistance rather than through peer review and discussion.

The errors or biases in these interactions may not be immediately catastrophic like a car accident, but their cumulative effect on healthcare outcomes, justice, and education could be equally devastating over time.

Just because error attribution is harder doesn't mean companies should avoid accountability. In fact, the moral and value decay potential rivals or exceeds that of more visible technologies. We're rightfully concerned about physical safety on our roads—shouldn't we be equally vigilant about the health of our information ecosystems and social institutions?

The Social Fabric at Risk

What we're seeing now is a subtle but profound shift:

• Individuals increasingly turning to AI rather than peers for information
• Decision-making becoming privatized rather than socially deliberated
• Knowledge validation happening through algorithms rather than communities
• Cultural transmission occurring through machines rather than intergenerational human contact

The risk isn't just about getting bad information—it's about atrophying our social thinking muscles. When we outsource thinking to machines rather than engaging with other humans, we lose the productive friction that generates new ideas, the emotional connection that builds trust, and the shared context that creates meaning.

Overindexing on Data

This connects to a broader problem of overindexing on data in organizational decision-making:

1. We mistake data volume for insight - Collecting massive amounts of information doesn't automatically translate to better decisions
2. We create false precision - Numbers can create an illusion of certainty even when based on flawed assumptions
3. We ignore unmeasurable factors - Things like community trust, organizational culture, or human dignity don't easily translate into metrics
4. We abdicate responsibility - Decision-makers sometimes hide behind "what the data tells us" rather than acknowledging subjective judgments

A Better Path Forward

The solution isn't rejecting technology but reframing its role. We need systems that:

• Augment human collaboration rather than replace it
• Make the limitations of AI transparent to users
• Value qualitative insights alongside quantitative data
• Preserve spaces for human deliberation and connection
• Balance efficiency with maintaining social capital

As leaders, we must ask not just "How can AI make us more efficient?" but "How can AI strengthen rather than erode the human connections that make our organizations and society function?"

The technologies that will truly advance humanity aren't those that isolate us with machines, but those that enhance our ability to think, create, and solve problems together.