by JL Ecochard
We approach new technology with fear and awe. Always have. Always will.
The problem is not the fear. The problem is that we experience fear through the cognitive lens of the present. We see what the technology looks like right now, with the mental frameworks in use in the present. We cannot see what it will become, or what it will do to us on the way there because we have not yet developed these future mindsets.
Our distant ancestors looked at a thunderstorm and thought it was alive. Not stupid. They had no framework for meteorology. The fear was entirely rational. Lightning kills. The thunder shakes the ground. Something that powerful had to be sentient, had to be angry, had to be appeased. The threat they imagined was not the actual threat. The actual threat, the atmospheric pressure system, the drought or flood that followed, the crop failure three seasons later, was invisible to them because they were using the only framework they had.
We are doing the same with AI. The fears dominating the conversation, such as job displacement, hallucinations, or bias, are real. They are also the visible risks, the ones our current cognitive framework can process. The toxic artifacts we cannot yet see are the ones that will shape us in ways we will only understand after the fact, when they have already hurt and polluted.
Every major technology disrupts the social infrastructure around it in ways nobody predicted from the tool itself. The steam engine did not just change how goods were made. It changed who was valued in the economy, which changed who was chosen, which changed how entire categories of people understood their place in the world. The social fractures that followed, the displacement, the identity crises, the movements that emerged to fill the vacuum, were not in anyone's forecast when Watt filed his patent. To anchor the point, the anxiety of the time was that the human body was not designed to survive speeds over 30 or 50 miles per hour. Fear was that blood would boil at speed faster than the fastest horse could ride. Likewise, we did not see brains' white matter changes coming from the smartphone. We will not see AI's equivalent in advance.
What we can do, the only honest practical response, is keep the possibility of invisible harm in our peripheral vision as we build and deploy. Not as paralysis. As humility. The goal is to see it a little earlier than after the fact. Not to prevent what we cannot predict. To notice sooner.
That is the argument. Not that AI is dangerous. Not that we should wait. That we should build with one eye on what we cannot yet see.
Not optimizing. Satisficing. Once a tool is good enough to rely on, we stop demanding better. We settle. We routinize. We stop questioning. The threshold for “good enough” drops over time. The scrutiny that accompanied the first use fades by the tenth.
Robert Cialdini spent his career mapping the mechanics of persuasion.² Reciprocity. Commitment. Social proof. Scarcity. Authority. Liking. Six principles that bypass deliberate judgment and activate automatic compliance. Not because we are weak. Because we evolved in a world where fast pattern matching kept us alive. Deliberate reasoning is expensive. Shortcuts are efficient. Most of the time they work.
AI is the carb-fat combination of cognition. Neither carbs nor fat alone are irresistible. Together they bypass the brake system entirely. A tool that always responds. Never judges. Always has an answer. Is frictionless, responsive, and perpetually available. The human brain did not evolve defences against that combination. We developed cognitive shortcuts for scarcity, not abundance. For friction, not frictionlessness.
This is not a reason to avoid AI. It is a reason to design its governance before the satisficing sets in. Because once it does, the window closes.
What this means practically: Build your AI governance frameworks now, before the tool feels normal. Because the moment it feels normal, nobody will question it. That is when the 23% fail rate stops being alarming and starts being acceptable. That is when the human who goes to jail stops checking.
Nobody voted for surveillance capitalism as the default business model of the internet. Nobody voted for attention extraction as the design principle of platforms their children now use twelve hours a day. Those defaults were set by people moving fastest while everyone else was still amazed.
The AI equivalent is being set right now. In board rooms. In grant cycles. In procurement decisions. In sessions like the one I ran at the Good Tech Summit last month. By people who mostly do not know they are answering questions that will shape what AI does in humanitarian contexts for the next decade.
Whose data trains the models. Whose problems get prioritized. Who owns the infrastructure. Who can audit the decisions. Who bears the cost when it fails. Who is the human who goes to jail.
These are not technical questions. They are governance questions. And the window to answer them intentionally, before the marvel phase passes and nobody is paying attention, is not years. It is months.
What this means practically: Your organization’s AI decisions in 2025 and 2026 are infrastructure decisions. Treat them that way. What you normalize now, you will not unnormalize later.
We are human in part because we have an oversized brain. That is not metaphor, the ratio of our brain to body size is anomalous among primates. That cognitive capacity is what made us the dominant species on earth, for better and for worse.
The industrial revolution used machines to compensate for weaker muscles. We cannot uproot a tree with our hands. With a digger, we can. We cannot fly between continents. With an airplane, we can. The machine extended what the body could not do.
But the machine did not just extend capability. It disrupted the hierarchy that physical strength had maintained for millennia. The burly man, whose muscle had conferred status, survival advantage, and reproductive selection, became less economically essential. The industrial economy valued a different kind of productivity. The social infrastructure built around physical dominance began to erode.
We did not see long term social ills coming from the steam engine. The displacement of physical strength as the primary currency of male status left a vacuum. Various forces rushed to fill it. Some of them became toxic, such as the manosphere. Not because the machine was malicious. Because the unintended consequences of technology on social infrastructure are always invisible until they are not.
Now we are using AI as a cognitive tool. To extend and augment the very trait that made us dominant. And we do not know what that will do to us.
We got a preview with social media. Studies published in Nature Communications and Scientific Reports, drawing on longitudinal neuroimaging data from adolescents, found significant associations between high screen time and structural changes in white matter, that is the neural architecture responsible for language, literacy, and emotional regulation.³ We handed children a dopamine machine and discovered years later it was rewiring their brains. We did not know. We could not have known. And then we did know, and kept going anyway – satisficed.
AI is a more intimate cognitive intervention than social media. It is not just competing for attention. It is substituting for cognition. For the generation growing up with it, the boundary between their thinking and the tool’s output will be genuinely unclear. That is not dystopia. That is the logical endpoint of a cognitive prosthetic that works.
We do not know what it will do. That is the honest statement. And it is reason enough to govern it carefully, deploy it intentionally, and refuse to let the marvel phase set defaults that we cannot undo.
Not panic. Not wait.
Name the AI decisions your organization is making right now that will be hard to reverse. The vendor you are locking into. The workflow you are rebuilding around a tool. The skill you are allowing to atrophy because the AI does it faster. Those are the decisions that matter. Not the acceptable use policy. Not just the ethics framework, the habits you are creating, the culture you are seeding.
The practical test: if the AI went away tomorrow, what would your organization no longer know how to do? That is what you are outsourcing. Decide deliberately whether that is acceptable.
Build governance before satisficing sets in. Include the people closest to the consequence in every design decision. Fund the infrastructure like roads. Name the human who goes to jail.
The window is open. It will not stay open.
JL Ecochard is Chief Innovation Officer at NetHope and co-founder of its Center for the Digital Nonprofit.
This is Part 4 of a four-part series developed from his session “When Every Hour Counts” started at the Good Tech Summit, April 2026. Part 1: Fund it like roads, not meals. Part 2: Nobody Goes to Jail. Part 3: Local.
Notes
¹ Herbert A. Simon, The Sciences of the Artificial (MIT Press, 1969). Nobel Prize in Economics, 1978. His concept of satisficing — choosing a solution that is good enough rather than optimal — describes how humans and organizations make decisions under complexity and constraint.
² Robert B. Cialdini, Influence: The Psychology of Persuasion (Harper Business, 1984). Six principles of persuasion — reciprocity, commitment, social proof, authority, liking, scarcity — that activate automatic compliance rather than deliberate judgment.
³ Crone, E.A. & Konijn, E.A. (2018). Media use and brain development during adolescence. Nature Communications, 9, 588. See also: Long-term impact of digital media on brain development in children. Scientific Reports (2024). Research consistently finds longitudinal associations between high screen time and structural changes in white matter tracts responsible for language, literacy, and emotional regulation, particularly in children and adolescents.
