Before we even ask if a machine will become conscious, we can observe what happens to our attention when we entrust it with some of the work. A suggested answer, a ready-made summary, a decision sorted by priority: each aid modifies the path through which we arrive at a result. The question of coevolution between humans and artificial intelligence begins here, in repeated gestures that seem only comfortable.

To speak of coevolution, in this context, is to formulate a question about mutual change. We train systems on human materials, then use their responses to produce new materials and orient ourselves. From this circuit does not automatically descend a new species, nor biological fusion with a machine. However, a concrete cultural problem arises: how much of our thinking becomes dependent on the technical environment we have built up?

The results we can observe

A job survey offers a first fixed point. Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied the introduction of a generative assistant among 5,172 customer service operators. In the version published in 2025, productivity, measured through the problems solved every hour, increases by an average of 15%. The benefits are unevenly distributed: the less experienced improve more. The study concerns a specific activity and enterprise; it does not measure the fate of world employment. Generative AI at Work, Quarterly Journal of Economics.

These data suggest an interesting possibility: a tool can make skills that previously required experience more accessible. The next step in the analysis is to ask what conditions transform care into learning. Who uses the system understands the reasons for the response? Can you recognize an exception? Does it maintain a space to develop knowledge of its own? The number of practices concluded does not alone answer these questions.

A second research introduces a different tension. In an experiment on short story writing, Anil Doshi and Oliver Hauser found that access to AI-generated ideas improves individual creativity assessments, especially for some participants. At the same time, assisted accounts are more similar to each other. The result concerns that experimental context, but opens up a question about cultural diversity. Science Advances, 2024.

We can each produce a more effective text and, collectively, inhabit a more uniform landscape. It is a consequence that deserves observation even where the success of the single seems evident: what variety of voices remains available when many travel shortcuts suggested by the same instruments?

Comfort has a direction

Let's imagine an editorial staff using a template to search for a topic within hundreds of transcripts. The gain is real: dispersed materials become searchable. But the initial question, the categories of research and the criteria by which we evaluate the results continue to orient what we see. If we only ask for confirmations, even a very rich archive can become a surface that always returns the same image to us.

Addiction can then be thought of as a loss of the ability to rephrase a question. When we always accept the first summary, who decides which detail is secondary? When a proposal arrives already ordered, do we still have time to explore what's left out? These are questions about our behavior, before the performance of the software.

The UNESCO Recommendation on the Ethics of AI addresses the problem of accountability: the use of systems does not eliminate that of people and organizations. Paragraphs 35 and 36 call for human supervision and determination, including when choosing to entrust the system with limited control. It is a normative position, not a demonstration that every application already achieves this balance. UNESCO, text of the Recommendation.

A Relationship to Choose

The possibility of an integrated mind between human and artificial remains a hypothesis to be defined and discussed. To evaluate it, we need explicit criteria: what do we mean by integration, what transformations do we observe and what consequences can we measure? An image of the future becomes more useful when it allows you to formulate verifiable questions.

In the present we can already exercise a choice: use AI to open alternatives, control their outcomes, preserve activities in which judgment matures through experience. Even silence before an answer has a value, because it leaves room for an intuition that is still formless. Technological transformation acquires a human direction when we know how to recognize what we want to entrust to it and what we want to continue to respond to.