The Room Where AI Gets Built

Imagine the team meeting where a major AI system's training objectives are decided. The people around the table have PhDs in mathematics, computer science, and statistics. They are exceptional at what they do. They are trying to solve hard technical problems under competitive pressure with significant resource constraints. Now ask: who in that room can articulate what Dostoevsky understood about how guilt reshapes a person's experience of time? Who has read seriously in the history of colonial thought and its relationship to extractive economics? Who can explain what Aristotle meant by phronesis and why it differs from knowledge? Who has spent years in conversation with people who will be affected by what the system does? This is not a critique of the engineers in the room. They are doing exactly what their training prepared them for. The critique is of a process that treats human-facing technology as a purely technical problem.

What Disciplines Are Actually For

The academic disciplines that AI development currently treats as decorative — philosophy, literature, history, anthropology, the arts — were developed to address exactly the questions that matter most in AI design. Philosophy developed systematic tools for thinking about values, consistency, and the conditions under which reasoning is reliable or fails. Literature spent centuries modeling the interior experience of human beings with enough precision to be persuasive across cultures and eras. History offers the most extensive database of what happens when powerful technologies and institutions encounter human nature at scale. Anthropology mapped the actual diversity of human values and social arrangements, resisting the temptation to treat one culture's norms as universal. These are not decorative additions to a technical project. They are the tools most suited to the hardest problems in that project — and they are largely absent from it.

The Tangent: Artists Have Been Here Before

Writers of speculative fiction were working through the implications of artificial intelligence decades before it became technically feasible. The questions they explored — what it means to create something that thinks, what responsibilities follow from that creation, how intelligence and consciousness relate, what happens when systems optimized for one goal encounter the complexity of a world they were not designed for — are the questions that urgently need answers now. Isaac Asimov's robot stories were not predictive in their specifics, but they were prescient in their structure. The Three Laws of Robotics fail in almost every story they appear in, not because the laws are badly written but because Asimov was demonstrating that any attempt to reduce complex moral requirements to a finite set of rules will encounter situations the rules were not designed for. That is a philosophical argument embedded in narrative form, and it is more sophisticated than most real AI ethics frameworks produced in the past decade.

What Happens Without This Input

The consequences of building AI without humanistic input are already visible in specific failures. Content moderation systems trained without genuine understanding of cultural context suppress political speech in some communities while missing hate speech in others. Automated hiring systems trained on historical data systematically disadvantage groups that were historically excluded from positions. Recommendation algorithms optimized for engagement metrics cause documented psychological harm to specific populations, particularly adolescents. Research from Princeton University's Center for Information Technology Policy examining AI system failures across sectors has found that the majority of consequential AI failures involve not technical malfunctions but value misalignment — systems working as designed but producing outcomes that designers should have anticipated and prevented. The designers did not anticipate them because they lacked the interdisciplinary training to ask the right questions before deployment.

The Structural Problem With How AI Teams Form

The incentive structures that shape AI development teams do not currently reward humanistic input. Publications in top venues reward technical novelty. Compensation structures favor people with skills in short supply. Timelines favor shipping working systems over conducting the kind of slow, careful inquiry that historical, philosophical, and anthropological analysis requires. Research from MIT Sloan Management Review on interdisciplinary team composition in technology companies has documented a consistent pattern: teams with diverse disciplinary composition produce innovations that perform better in the market and generate fewer harmful externalities, but they take longer to produce results and require more active management. Under competitive pressure, companies systematically under-invest in the kind of composition that produces better long-term outcomes.

The Practical Ask

This does not require dismantling existing AI development processes. It requires adding serious voices from disciplines that have spent centuries on the relevant problems. It requires creating conditions where those voices can be heard — not as compliance requirements or PR exercises but as genuine intellectual contributors to design decisions. The people who built the internet did not ask humanists what kind of social environment would emerge from global always-on connectivity. The consequences of that omission are well documented. The people building AI systems are in a position to make a different choice. Whether they will is a different question.