The biases of trust in machines

We proclaim our attachment to human autonomy in the face of machines, but our everyday behaviors reveal a very different reality: we willingly delegate our decisions to them.

The Paradox of Our Voluntary Submission

Discourses on artificial intelligence abound with warnings: we must not lose our sovereignty, machines must not decide for us, our freedom of thought must remain intact. Yet, for all the people I interview about their concrete experiences, I observe quite a different picture: the car that brakes automatically and avoids an accident becomes a precious ally. Meeting summaries generated by ChatGPT free up a considerable amount of time. Almost everyone uses GPS for navigation. And so on. We integrate these tools with disconcerting ease, quickly giving them a central place in our lives.

This contradiction between discourse and practice reveals what Sherry Turkle calls the “ELIZA effect” in her book Alone Together (2011): our tendency to attribute intelligence and intentionality to machines that do not necessarily possess them. We know intellectually that they are algorithms, but we act as if they were wise advisors — all the more so because these advisors are not human.

Philosopher Luciano Floridi speaks of a “fourth revolution” in which we become “inforgs,” informational organisms evolving within the infosphere. This transformation occurs not by constraint, but through a gradual and voluntary adherence to systems that appear more reliable than our own judgment or that of our peers.

The Illusion of Technological Objectivity

GPS offers the most revealing example of our willing surrender. We follow its instructions with remarkable docility, even when our own experience tells us there is a better route. I have known my neighborhood for twenty years; I know perfectly well that GPS is making me take an absurd detour, and yet I keep following that synthetic voice. Why such passivity, when we are the ones actually holding the steering wheel?

This blind trust stems from what I call the “algorithmic authority bias”. We attribute to the machine a superior rationality, assuming it holds information that escapes us. As Cathy O’Neil analyzes in Weapons of Math Destruction (2018), we tend to sanctify algorithms as if they were neutral and omniscient, forgetting that they are designed by humans with their own biases and goals — and that even when they learn on their own, as is the case with deep-learning-based AI, such as generative AI, they are not gods of ubiquitous knowledge. Biases, errors, and hallucinations remain abundant.

In taxis, this GPS phenomenon sometimes reaches its climax. Passengers may demand that the driver follow the GPS to the letter, preferring the apparent objectivity of the machine to the expertise of a professional who knows their city. They fear human deceit but choose to ignore potential algorithmic manipulations. In the case of Uber, for instance, the algorithm is anything but neutral: it aims to maximize the platform’s profits. A systematic bias slightly lengthening each route could generate millions in additional revenue. Despite this obvious fact, we persist in trusting the algorithm more than the flesh-and-blood driver — and this is far more widespread inside taxis than we tend to think.

The Consented Dehumanization of Our Interactions

Taxi drivers, whether independent or working for platforms, often face the wrath of passengers whenever they dare to take a human initiative — a shortcut based on their experience. Confronted with such hostility, they relinquish their expertise and become docile executors of the algorithm. Thus we create a vicious circle in which humans dehumanize themselves to conform to social expectations that glorify machine-like pseudo-objectivity.

This process echoes what sociologist Hartmut Rosa describes in Resonance: A Sociology of Our Relationship to the World (2018): we may gradually lose our capacity for resonance with the world and with others, preferring predictable interactions with digital interfaces. The machine becomes a reassuring mediator that spares us the complexity and unpredictability of authentic human relationships (even though the machine can make gross errors — which we forgive far more easily than we would a human).

I do not claim that machines are useless; human–machine collaboration holds extraordinary potential when it remains in service of our humanity. But in that taxi, instead of distrusting the person driving us, we could choose to truly meet them. We could recognize together that a machine is interposed between us, and affirm our freedom as human beings capable of dialogue and shared decision-making — with the machine remaining a tool, not a form of dictatorship lodged at the heart of our intimacy, chosen by ourselves for ourselves.

The Erosion of Interpersonal Trust

The core of the problem lies in the crumbling of trust between human beings. This crisis of trust makes the question of empathy all the more urgent in our time of anthropological transformation. Cultivating connection, reaffirming our humanity, rediscovering a living humanism — that, to me, seems essential.

As Yuval Noah Harari notes in 21 Lessons for the 21^st Century (2018), we risk becoming a “useless class,” not because machines will replace us, but because we will have lost confidence in our own judgment and in that of our fellow humans. This self-fulfilling prophecy drives us to systematically prefer algorithmic authority to human expertise and creativity.

The ultimate paradox is already looming: future autonomous taxis, equipped with synthetic voices capable of consulting us about our route preferences, will probably inspire more trust than current human drivers. We will appreciate their programmed pseudo-humanistic behavior, while among ourselves we adopt increasingly machinic, procedural, and disembodied ways of interacting.

Reinvesting in Our Humanity in the Algorithmic Age

This drift is not the machines’ fault but our collective responsibility, rooted in our daily gestures. We cannot prevent their omnipresence, but we can choose to cultivate our human connections more intensely. Machines free us from time and effort; let us seize this opportunity to reinvest in what defines our humanity: empathy, spontaneous creativity, and contextual judgment.

Martin Buber, in I and Thou (1923), distinguished between two fundamental modes of relation: the “I–It” (instrumental) and the “I–Thou” (authentic). Our challenge is not to turn all our relationships into “I–It” interactions mediated by algorithms, but to preserve and enrich true “I–Thou” encounters. No machine has the power to prevent this; it is our responsibility — and it remains entirely within our reach.

The issue is not to reject machines, but to keep them in their rightful place as tools. As Jürgen Habermas writes, we must preserve our “lifeworld” (Lebenswelt) against colonization by technical systems. This requires constant vigilance and a deliberate effort to keep alive our autonomous judgment, interpersonal trust, and authentic interaction. Machines can enhance our capacities, but they must never replace our fundamental humanity — we must instead use them as tools that contribute to it.