What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
AI can help people make better decisions when its advice is accurate, but it can also pull judgment in the wrong direction when its answer is wrong or biased. Studies show a more conditional picture than the headline claim: speed and confidence can coexist with overreliance, and the evidence does not establish that newer or more capable models generally make people worse at judging.
Does using AI make your decisions worse?
It can, in some tasks and circumstances—but not simply because the model is more capable. The important distinction is between what a system can do and what happens when a person relies on its output. Accuracy, bias, task design, interface cues, and the user’s response all affect the result.
As an Amazon Associate I earn from qualifying purchases.
In two online experiments on LLM-based search, participants completed tasks faster than people using traditional search. Their decision accuracy was comparable when the LLM was correct, but they overrelied on incorrect information when it erred. As the study authors put it, “While decision accuracy was comparable when the LLM was correct, users overrelied on incorrect information when the model erred.” Speed, then, is not the same thing as better judgment.
That distinction matters because a tool can save time or improve performance on a task without making every decision more reliable. The studies cover specific experimental settings; they do not show that people’s judgment always declines with AI use.
#1 Best Overall
How can AI influence human judgment?
Accurate advice can help; biased advice can amplify bias
Experiments by Moshe Glickman and Tali Sharot, involving 1,401 participants, found that accurate AI improved judgments in the study’s task paradigm. Biased AI, however, could amplify biases in human perceptual, emotional, and social judgments. Participants were often unaware of how much the AI had influenced them.
Glickman and Sharot describe the study-specific finding this way: “The current study reveals that biased algorithms not only produce biased evaluations, but substantially amplify such biases in human judgements, creating a feedback loop.” This is evidence about the tasks they tested, not proof of equivalent effects in every workplace, medical, or civic decision.
People may accept incorrect recommendations
In preregistered personnel-selection experiments reported in Scientific Reports in 2024, incorrect AI advice harmed task performance because participants did not reliably dismiss it. Making predictions more explainable did not reliably eliminate that overreliance in those experiments. A visible rationale can be something to examine, but it is not proof that a recommendation is correct.
What safeguards help—and what they cannot guarantee
Check the answer independently
For consequential decisions, verify factual claims using an authoritative source independent of the AI. This follows from evidence that people can overrely on erroneous output and from the OECD’s discussion of automation bias: accepting incorrect AI results or failing to cross-check them. The OECD report is policy synthesis, not a controlled experiment or a prevalence estimate for all AI users.
Form an initial view before consulting AI
When the stakes justify it, write down your decision criteria or initial judgment before reading a recommendation. That gives you a point of comparison and may make it easier to notice when an answer is steering you. This is a prudent practice inferred from overreliance findings, not a universally tested intervention.
Treat interface cues and second opinions as aids, not fixes
In one LLM-search experiment, color-coded highlighting helped participants identify errors and improved decision accuracy without changing other outcomes. A 2024 ACM study found that adding a second opinion—whether from a peer or another AI—could reduce overreliance while also increasing underreliance. Neither safeguard is a guarantee: one was tested in a particular search task, and the other could shift trust in either direction.
How should you assess an AI-assisted decision?
Rather than judging a tool only by its apparent reasoning ability or speed, ask how it affects the decision in front of you. These questions synthesize outcomes considered in different studies; they are not a validated scoring system.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- Can you verify the output? Look for claims that can be checked against an independent, authoritative source.
- How does the process perform when the AI is wrong? A strong result when the recommendation is correct does not reveal whether users catch errors.
- Does the interface make errors or uncertainty visible? A cue may help in a specific setting, but should not substitute for checking.
- Is your own judgment elicited? An initial view or explicit criteria give you something to compare with the recommendation.
- Does a safeguard improve calibration in both directions? Reducing overreliance is not enough if people then dismiss useful advice too often.
Does a better model mean better human decisions?
Not necessarily. A 2024 Communications Psychology paper compared human and language-model reasoning on variants of the Cognitive Reflection Test and the Linda/Bill conjunction problem, and discussed how prompting can affect model performance. Such comparisons show that measured reasoning depends on the task and prompt; they do not test whether improving a deployed model makes its users worse judges.
The cited human–AI studies examine different tasks, interfaces, and outcomes. They do not establish a general causal trend across successive model generations, nor do they settle whether routine AI use changes people’s independent judgment over the long term. There is also no single comparable effect-size figure that responsibly summarizes the claim that “AI makes human judgment worse.” The defensible conclusion is narrower: AI can help or mislead, and the impact depends on the quality of its advice and how people use it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
