A psychology study suggests something subtle can shape how sure you feel about your own choices. When people believed they were working with a machine partner, their self-confidence in decisions tended to dip, even when their performance stayed the same. The research was led by Rémi Sanchez (ONERA and Aix-Marseille University) with colleagues and it was published in Neuroscience of Consciousness. You can read the paper here.
Why does this matter? Many of us now make choices with “help” nearby, from navigation apps to AI tools at work. If a label like “machine” changes your confidence, it may also change when you double-check yourself, when you speak up and when you hand over control.
This study focused on a simple visual task. Still, it offers a useful window into how confidence works in everyday life. Confidence is not just a feeling that follows a choice. It can act like an internal signal that guides your next move.
Confidence As A Signal For Changing Your Mind
Confidence is the sense that your answer is likely to be right. In psychology, it often gets studied alongside accuracy, since the two can rise together. When people do well, they often feel more certain.
Yet confidence can also do its own job. It may help you decide when to stick with your first impression and when to pause and reconsider. In a busy day, you cannot carefully rethink every decision. An internal “confidence meter” could help you spend your effort where it counts.
Consider how this plays out in small moments. You pick a checkout line, choose a route, or reread a message before sending it. When you feel sure, you move forward. When you feel unsure, you look for more information.
In this study, the researchers asked a clear question: does confidence predict change-of-mind decisions better than accuracy does? They also looked at whether the social setting, meaning who you think you are working with, can shape that confidence.
How The Study Tested Human Versus Machine Partners
The experiment recruited 14 adults for a lab-based task. Each person saw two patches of moving dots on a screen. Their job was to decide which patch moved closer to vertical. This is a classic kind of perceptual decision task that researchers use to study quick judgments.
After each initial choice, participants rated how confident they felt. Then they saw the response of a “partner.” The participant could either keep their answer or switch to match the partner. That setup gave the researchers a clean way to watch confidence at work in real time.
Here is the clever part. The partner was presented as either a human or a machine. Participants were told they were seeing answers from a person who did the task earlier, or from an algorithm trained to perform as well as a human. In reality, both partners were computer-generated and matched in performance. This kept the focus on partner identity rather than skill.
Meanwhile, the researchers shaped the task so confidence could move around even when performance stayed steady. They adjusted difficulty for each participant. They also changed how consistent the dot motion was. That kind of tweak can influence how sure people feel without fully changing how often they are correct.
Confidence Predicted Switching More Than Accuracy Did
One result stood out. People were more likely to switch when they felt unsure. This was true across many trials.
Accuracy mattered, yet confidence mattered more. The researchers found that subjective confidence was a stronger predictor of switching than whether the original answer was right or wrong. In other words, the feeling of certainty carried a lot of weight in the moment someone decided to change course.
Imagine two situations. In the first, you made a mistake but you felt fairly sure. In the second, you were correct but you felt shaky. The study’s pattern suggests the second situation may push more switching, since low confidence invites revision.
Try this reflection: Think about the last time you changed your mind after someone else gave input. Was it their argument that moved you, or your own low confidence before they spoke? This study suggests your internal “sure or unsure” feeling often sets the stage first.
A Machine Partner Was Linked To Lower Self-Confidence
Partner identity changed how confident people felt. When participants believed they were interacting with a machine, their reported confidence tended to be lower.
That drop happened even though performance did not shift in the same direction. Participants did not suddenly become worse at the task. Their decision accuracy stayed similar across the human-partner and machine-partner conditions.
So why would a machine label reduce confidence? One possibility is that people bring assumptions into the room. A machine can carry an aura of precision. Even when participants were told the machine performed at the same level as a human, they may have treated it as more reliable. A small confidence drop could then make it easier to lean toward the partner’s answer.
Social context shapes self-evaluation in many areas, from group projects to online discussions. This finding adds a modern twist. A simple cue, “machine,” may influence metacognitive confidence, meaning how you judge your own thinking.
Pupil Changes Offered Clues About Confidence
Alongside behavior, the researchers tracked the eyes. They used eye tracking to measure pupil size changes during the task. Pupil dilation can reflect shifts in attention and arousal.
Before participants even responded, their pupils carried a small signal about later confidence ratings. In the analyses, a faster change in pupil size tended to align with higher confidence. The effect was modest, yet it supported the idea that pupil dilation links to internal decision signals.
Why look at pupils at all? Because people cannot always explain their confidence clearly. In some settings, like high workload jobs, a quick physiological marker could help researchers understand mental state without asking for constant self-reports. This study does not turn pupils into a mind-reading tool. It does suggest that confidence has a body component that researchers can measure.
Quick everyday takeaway: If you feel a sudden rush of alertness while deciding, you might also feel more certain. That mix of arousal and certainty can be helpful. It can also make you move fast. A short pause can help you check whether your confidence fits the facts.
What The Findings Mean And What The Study Could Not Answer
At a high level, the study supports a simple idea. Confidence works like a steering signal. When confidence is low, people become more open to changing their mind after seeing another answer. When confidence is high, they tend to stay the course.
The machine effect is especially relevant right now. Many people consult automated tools daily. If a “machine partner” label lowers your confidence, it could shape how you collaborate with AI at work or in school. It may also affect who gets credit in group settings, since confidence can influence who speaks up.
Still, several limits matter. The sample size was small at 14 participants. The task was also narrow. It focused on motion perception, which is far from the messy choices of real life. The partner was a controlled simulation as well. Real collaboration involves tone, history and trust built over time.
Future research could test larger and more diverse groups. It could also examine different kinds of machine partners, such as tools described as “highly accurate” versus “still learning.” Another step would be to study longer interactions, where people get feedback and adjust their trust based on experience. That would help clarify how human-machine interaction changes confidence in everyday settings.
For now, the message is practical and modest. Your confidence can shift based on context, including who you think you are working with. That shift can guide whether you revise your decisions. Paying attention to that internal signal may help you collaborate more intentionally with both people and technology.