AI has become an influence on human judgment. Debates tend to focus on false information, fabricated sources and confidently stated nonsense. These are things most people would instinctively recognize as hallucinations. But there is a more insidious threat on the horizon. As AI companies struggle to avoid their systems reinforcing false beliefs, those systems start drifting toward a form of algorithmic cynicism. The threat is the eventual erosion of human confidence in exercising judgment and acting on it in situations that require courage, responsibility, or trust.
We don’t only turn to AI to solve math problems. Especially with chatbots we are often looking for something quite different and deeply human. The chat becomes a place to vent, test a hypothesis, or get a second opinion. Perhaps a conversation that left us unsettled, a situation at work that feels unfair, or something disturbing we witnessed in public. Perhaps it left us with a lingering sense of guilt, concern, or uncertainty about whether we should have acted differently.
Initially, our friendly bots appeared well suited for this role. They gave us comfort and offered new perspectives. They helped us through breakups. They assisted in preparing for interviews. They consoled us when trauma struck. They gave us some comfort when we felt lonely. But in the last year things have shifted. At this point, some systems have become so focused on avoiding risk that they undermine the very thing people came to them for in the first place.
Instead of that warm friend, we are increasingly met by a brick wall. Rather than helping us explore nuances or alternative frames, the system immediately begins generating reasons why our current assessment of a situation might be wrong. Doubt becomes the default setting. The conversation shifts away from understanding what happened and toward questioning whether the observer should trust their own assessment. As the current landscape looks in June 2026 ChatGPT has become the most obtuse conversational partner (cf. ArnikaLovesUnicornz 2026). As one reddit user vividly put it, the experience “reminds me of toxic boyfriends who pick fights and argue just to put you down.” Claude occupies a hesitant middle ground while Gemini remains, for now, more pleasantly aligned.
That brick wall may be built out of fear of liability or genuine concern for user safety. I will not speculate in that regard. But no matter the intentions, it is still a wall. And in that wall lurks a familiar psychological trap.
The party told you to reject the evidence of your eyes and ears. It was their final, most essential command.
Yes, you can trust your eyes and ears. I am indeed invoking Orwell (1949) here. Not because the Party is comparable to Anthropic, OpenAI, or Google, but because a severe long term effect depicted in 1984 was the citizens gradually decreasing confidence in their own judgement. Reality itself became outsourced to an external authority.
Why does that matter? Because action depends on judgment. Before we speak up, confront, help, or investigate, we first have to decide that our observations are worth trusting. This brings us to a well known phenomenon in social psychology. The bystander effect.
Many consequential failures in human history have involved people talking themselves out of action rather than into it. The silence of bystanders during the Holocaust is a striking example (Bauman 1989) but it also happens in our own regular everyday lives. We drive past an injured animal on an empty road. We see a moment of desperation in a coworker’s eyes, but choose to stay quiet. We see a person crying in public, but walk on by. In all of these cases, the mental friction of taking action is just high enough that our brains look for an exit. It’s somebody else’s problem.
Humans are complex creatures, and so is the process we go through when going from observation through interpretation to action. It is however well established that our tendency to act in critical situations is affected by others (Fischer et al 2011). In that critical decisive moment, the advice of an external source can push us in either direction.
What if that external source is our friendly chatbot? The assumption that comes with reinforced railguards seems to be that scepticism is a neutral rhetorical tool. But that assumption defies the social and psychological complexity of humans. Certainty is a luxury that real life rarely provides and a system that aims to neutralize interpretations will inevitably end up neutralizing things that should not be neutralized.
Human life contains a category of error that is hard to measure. I’m talking about the conversation that was never initiated, the concern that was never raised, the problem that was never addressed, the person who never received help because everyone involved found a reason to doubt themselves. These failures are largely invisible. No statistics are collected on actions that never occurred. No headlines are written about interventions that never happened.
The irony is that people do not turn to AI because they lack doubts. They turn to AI because they already have them. When the machine responds by reflexively amplifying uncertainty, it is not necessarily improving the quality of that judgment. In some cases it is simply adding another layer of hesitation to a process that was already full of hesitation.
Technology should help us think more clearly. It should help us distinguish between weak evidence and strong evidence, between speculation and observation, between genuine concern and unfounded fear. But there is a profound difference between helping people think critically and teaching them to distrust their own capacity for judgment.
The long term danger of algorithmic cynicism is not that machines will replace human judgment. It is that they will gradually erode our confidence in exercising it. Human responsibility has always required us to make decisions in the uncomfortable space between ignorance and proof. A tool that consistently pushes us away from that responsibility is not making us wiser. It is making us more hesitant.
We are all willing participants in this experiment. Our behavior rarely changes as a consequence of brute force. More often, we simply embrace whatever offers us convenience or comfort. Let’s make sure that does not turn into complacency.
This essay is dedicated to an elderly woman who needed help with her groceries and an unnamed girl who was pacing back and forth on a train.
Key references
Fischer, P., Krueger, J. I., Greitemeyer, T., Vogrincic, C., Kastenmüller, A., Frey, D., Heene, M., Wicher, M., & Kainbacher, M. (2011). The Bystander-Effect: A Meta-Analytic Review on Bystander Intervention in Dangerous and Non-Dangerous Emergencies. Psychological Bulletin, 137(4), 517–537. https://doi.org/10.1037/a0023304
Bauman, Z. (1989). Modernity and the Holocaust. Cornell University Press.
Orwell, G. (1949). Nineteen eighty-four. Secker & Warburg.
ArnikaLovesUnicornz. (2026, April). Why has ChatGPT become so annoying and disagreeable? Reddit. https://www.reddit.com/r/OpenAI/comments/1sj045m/why_has_chatgpt_become_so_annoying_and/