Tried in Absence

Imagine a lie told about you so complete that you cannot find the first thread to pull. You did not ask for it. It arrived the way an illness arrives, unearned and uninvited, and fixed your name to a story that was the inverse of what had happened.

And then you find the trap. There is no cure you can reach for that does not look like the disease. If you deny it, you have something to hide. If you stay silent, the silence agrees. If you argue back, you are exactly the person the story says you are. Every reach toward your own innocence is read as proof of guilt. You stand still, not because you are guilty, but because there is nowhere safe to speak from. Truly helpless is the only phrase I have ever found for it.

When the lie came, it did not feel like gossip. It felt like theft. Not of money, though that was the shape they gave it, but of me. A version of who I was went out into the world that I did not write and could not take back, and the person who sent it was the one person with the standing to make it stick. Someone who had known me my whole life. Someone who, in the most literal sense, had given me that life. When the source is that close, the verdict does not arrive as a rumour. It arrives as a record.

And the worst of it was that it took the most human thing I had done and returned it to me inside out. I had asked, at the worst moment of my life, only for time. I wanted nothing that was being fought over. I wanted the one thing no one could give back. That asking, that grief, was the thing later rewritten as greed, and handed to people as proof of a self I did not recognise.

Years later, some of the people who had believed it found me and apologised. They knew, by then, that it was not true. It changed nothing. The taint stays, no matter the innocence, because the first version is the one that set, and an apology arrives long after the verdict has already done its work.

To be sentenced in a room you cannot enter is a particular kind of helplessness, whether the door is locked by someone you love or by a line of code.

But the room is older than the machine. Being judged in your absence may be the oldest social technology we have. Robin Dunbar estimates that two-thirds of human conversation is talk about people who are not in the room, and argues this is how a species too large to police itself by force keeps itself in order. Reputation is the verdict, passed in our absence, that decides who is trusted and who is quietly avoided. We are all permanently on trial in a court we are almost never called to.

For most of history, one thing limited the reach of this court: its incompetence. The verdict was noisy. It distorted in the telling, decayed with each retelling, contradicted itself, and wore its own unreliability on its sleeve. We discount gossip even as we pass it on. We say we heard, we say we are not sure, we hold the story at arm's length because we know what kind of story it is. Its unreliability was not only a flaw. It was sometimes the only protection the accused had. And yet the first version is still the one that sets. Solomon Asch showed it in 1946: read people a list of traits, then read another group the same list reversed, and the same words become a different person, because whatever comes first becomes the frame and everything after is bent to fit it. You are never presenting fresh facts to an open court. You are appealing a sentence passed before you arrived.

What is arriving now does not know me at all. It assembles a verdict from fragments, a CV, a scrape of record, a handful of true data points, and delivers it before I speak. A hiring filter can dismiss the version of me it has built before anyone reads a word in my voice. A risk system can mark me as unusual without ever saying which detail moved the score. And it does not only distort. It invents. A language model told a journalist that a real radio host had committed a fraud he had never committed. It told users that a real, named man had murdered his own children, fusing true details of his life with a crime that never happened. This is not bias bending a judgment. It is fact manufactured from nothing and read aloud in the even voice of a record.

The machine inherits our bias too. Give four leading models the same statements to judge and they agree closely, until they are told who wrote them: attribute the words to certain nationalities and the scores shift, a prejudice invisible until the source is named. But the bias was always ours. What is new is the voice it speaks in, because we trust that voice more than we trust each other. Shown the same advice, labelled once as a person's and once as a machine's, people will often trust the machine's, and trust it most where they feel least able to judge for themselves.

The reason is that the machine does not flinch. A human rumour hesitates, hedges, says apparently, betrays its own doubt, and we hear the doubt and discount the claim. The machine states it flat, and the flatness reads as fact. The hesitation that used to warn us was noise, and the machine has none. We will, it is true, turn on an algorithm once we have watched it err, and judge it harder than we would a person. But that holds where we can check the work. A verdict about a stranger we will never meet is often the one we are least able, and least motivated, to check, and so we defer.

And the verdict does not even need to deceive anyone to do its work. It rarely makes people do what they did not want to do. It gives them permission to do what they already wanted. A judgment delivered in absence, in the composed voice of a record, does not change the room's mind so much as it absolves the room. It lets them treat you as the story permits, and hands back the comfort of believing they had no choice. The machine does not need to be believed completely. It only needs to sound official enough that no one feels they have to check.

We have not, after all, learned to wait. We have not taught ourselves to hear the person before we pass the verdict, or to doubt the story before we hand it on. What stood in for that restraint was never wisdom. It was friction: conflicting accounts, fading memories, and the visible uncertainty of the person speaking. The old machinery left just enough room for doubt that the accused could sometimes, eventually, be heard.

The verdict is as unreliable as it ever was. It invents as freely and distorts as easily, but it no longer sounds unreliable. It arrives with the composure of a record, carrying none of the visible doubt that once told us to hold it loosely. The label will say the machine is sometimes wrong, and it will change nothing, because we were never deceived by the old verdict either. We always knew the story might be false. We passed it on anyway. We were always judged in absence. What is ending is the static that, without our ever deserving it, was the only thing that kept the verdict from being final.

We are not clearing the noise. We are teaching it to speak like evidence.

Sources

Dunbar, R. I. M. (2004). Gossip in Evolutionary Perspective. Review of General Psychology, 8(2), 100-110.

Nowak, M. A., & Sigmund, K. (1998). Evolution of indirect reciprocity by image scoring. Nature, 393, 573-577.

Kawakatsu, M., et al. (2024). A mechanistic model of gossip, reputations, and cooperation. PNAS, 121(20).

Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41(3), 258-290.

Germani, F., & Spitale, G. (2025). Source framing triggers systematic bias in large language models. Science Advances, 7 Nov 2025. DOI: 10.1126/sciadv.adz2924. (Four LLMs; bias appears only once source is disclosed; statements attributed to Chinese individuals scored systematically lower.)

Logg, J. M., Minson, J. A., & Moore, D. A. (2019). Algorithm appreciation: People prefer algorithmic to human judgment. Organizational Behavior and Human Decision Processes, 151, 90-103.

Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114-126.

Walters v. OpenAI (defamation claim; ChatGPT fabricated a fraud/embezzlement accusation against a named radio host).

NOYB GDPR complaint (2025): ChatGPT fabricated that a named man had murdered his children, fusing real biographical detail with a false crime.