There is a curvature in the light, an imperceptible deviation in the path of photons that alters every landscape. It is not in the objects, but in the eye that receives them – or in the lens that should be transparent. This curvature is called bias, prejudice, distortion. It is the mental map we mistake for territory, the ancestral whisper that shapes our silences before words are even born.

The sample we believed to be representative was instead a family portrait. Entire populations erased from the data as if they did not exist, while the same voices – always the same ones – echoed in the algorithm’s chambers. The unwitting archivist building infinite libraries, all identical, where every volume repeated the same misprints. The man who looks for his keys not where he lost them, but only under the streetlamp because there is light. And the light – oh, the light – is itself a selection, an omission, a betrayal.

Data is no more neutral than water flowing in a stone bed: it takes the shape of the container, absorbs its flavor, its geological memory. Those datasets on which we train our digital oracles contain entire centuries of crystallized injustices, prejudices turned into statistics, discriminations become code. Machine learning invents nothing: it only amplifies what we already were, returning it to us multiplied a thousandfold, like a mirror that, instead of reflecting, distorts to the point of caricature.

And the human mind? That magnificent and fallacious organ that builds coherent narratives from chaotic fragments. Confirmation bias is not a flaw, but a survival strategy: we seek confirmations because total doubt would be an ocean in which to drown. We prefer to be right in a small world than wrong in a vast universe. We build comfortable cages and call them truth.

Institutions are these slow animals that crystallize individual biases into steel structures. They become invisible architectures guiding our steps along predetermined corridors, even when we believe we are choosing the direction. Prejudice is no longer a personal opinion, but the slope of the floor, the height of the doorframes, the arrangement of the doors.

Antidotes exist, of course. Representative samples that show us not only our own face but also the profiles we do not recognize. Debiasing techniques that are exercises in epistemological humility. External validation that is always a questioning of oneself. Transparency that is making the seams of the fabric visible, showing the points where the reasoning might unravel.

But perhaps bias reduction is not a technical matter, but an ethical and aesthetic one. It is the search for a broader, more inclusive beauty, one that does not exclude any form of life or thought from its composition. It is the daily exercise of looking beyond one’s visual angle, of imagining the experience of those on the other side of the mirror.

Bias remains the blind spot of every universe – personal, algorithmic, social. That zone of shadow that necessarily accompanies every light. Perhaps we cannot eliminate it completely, just as we cannot see our own gaze. But we can acknowledge it, measure it, compensate for it. And above all, we can remember that every truth we believe to be absolute already bears the marks of the viewpoint from which it was observed.

Perfection lies not in eliminating every distortion, but in recognizing it as a constitutive part of our way of being in the world. In accepting that we are finite creatures who aspire to the infinite, and that perhaps in this gap – between what we are and what we might see – resides our most human, imperfect, wonderful dignity.