Quick answer

To read other people more accurately, rely less on imagining what they think or feel and more on finding out.
Research on empathic accuracy shows that confidence in your read is a poor guide to whether it is correct, while real information from the other person improves accuracy.

Five ways to improve empathic accuracy:

  • Create an accuracy stake: make a prediction you will genuinely check afterwards.
  • Ask rather than imagine: replace perspective taking with perspective getting whenever you can.
  • Check your confidence: high confidence does not necessarily mean your interpretation is accurate.
  • Look for specific differences: consider how the other person’s incentives, history or situation differ from yours.
  • Check predictions against reality: compare what you expected with what the person actually reports.

The simplest way to get better at reading people is to stop treating your interpretation as the answer. Perspective getting, actually asking and checking, is more reliable than perspective taking alone.

You’re fairly confident you know what that raised eyebrow meant. You’re fairly confident you know why the room went quiet after you spoke. The trouble is, confidence and accuracy turn out to be almost entirely separate things when it comes to reading other people, and most of us have never actually checked whether our read was right, only whether it felt right.

That gap matters more for a leader than almost anyone else in the room, since so much of the job runs on correctly interpreting reactions you don’t get to ask about directly. The useful news is that accuracy here isn’t fixed. It’s a skill with a specific, well understood failure mode, and once you know what that failure mode is, it’s genuinely trainable.

What this actually buys you as a leader of yourself

Leaders who misread reactions don’t usually find out in the moment. They find out later, when a decision they thought landed well turns out to have quietly worried half the room, or when someone they assumed was fine turns out to have been disengaged for months. The cost of misreading people isn’t a single bad guess, it’s every decision built on top of that guess afterwards.

Getting better at this isn’t about becoming more sensitive in some vague, general sense. It’s about noticing the specific place your read of someone usually goes wrong, and building a habit that corrects for it, which is a much more achievable target than simply trying to be more empathetic.

Why it matters

Empathic accuracy, the ability to correctly infer what someone else is actually thinking or feeling, is measurable, and once researchers started measuring it directly, a few things became clear. People vary substantially in how accurate they are, accuracy doesn’t correlate well with how confident someone feels in their own read, and critically, accuracy can be trained rather than being a fixed trait some people simply have (Ickes, 1997).

Part of why we get it wrong has a specific, well documented mechanism. When people try to imagine another person’s perspective, they don’t start from a blank slate, they start from their own perspective and adjust away from it. The problem is the adjustment is usually too small. We anchor on what we ourselves would think or feel in that situation, then shift only slightly towards what we imagine the other person’s view might be, which means our read ends up closer to our own perspective than to theirs (Epley et al., 2004).

There’s an even sharper finding sitting underneath that one, and it’s genuinely uncomfortable. Across twenty five separate experiments, people instructed to imagine another person’s perspective were, on average, no more accurate than people given no such instruction at all, and in several of the experiments, imagining actually made accuracy worse while making people feel more confident in their answer (Eyal, Steffel and Epley, 2018). The advice to “put yourself in their shoes” is one of the most common pieces of leadership guidance there is. The research says imagining alone usually doesn’t help, and can quietly make things worse while feeling like it’s working.

What did help, across the same body of research, was genuinely different, having real information about the other person, through conversation, disclosure, or simply asking, rather than working from imagination alone. That’s the distinction the empathic accuracy paradigm is built on too, comparing a prediction against what someone actually reports feeling is how accuracy gets measured in the first place, and it’s a method you can borrow directly, not just something researchers do in a lab (Ickes, 1997).

Imagining a perspective versus getting one

Approach What it actually involves Why it tends to go wrong or right
Perspective taking Imagining how the other person sees the situation, without new information Starts from your own view and under-adjusts away from it, on average no more accurate than not doing it at all, and can increase confidence without increasing accuracy.
Perspective getting Actually finding out, through conversation, disclosure, or asking directly Introduces real information the imagining alone doesn’t have access to, which is what the accuracy actually depends on.

The uncomfortable implication is that the more effortful, more empathetic-feeling option, sitting and imagining someone else’s view carefully, doesn’t reliably beat doing nothing. The plainer option, finding out, is the one actually doing the work.

Five practices for reading people more accurately

Exercise 1: The accuracy stake

Overview: Egocentric bias in perspective taking measurably decreases when people have a genuine incentive to be accurate, not just an instruction to try harder or think more carefully (Epley et al., 2004).

Steps:

  1. Before an interaction where you’re about to form a read of someone, create a real stake, say your prediction out loud to a colleague, or write it down somewhere you’ll have to revisit.
  2. Make the prediction knowing it will actually be checked, not just privately guessed and forgotten.
  3. Follow through afterwards and genuinely check it against what happened.
  4. Notice whether having a stake changed how carefully you thought it through, compared with a private, consequence-free guess.
  5. Build this into situations that matter most, rather than only using it for low stakes practice.

Variations: A colleague willing to trade predictions with you, each checking the other’s, makes the stake easier to sustain than doing it alone.

Why it matters: This uses the one thing the original research found reliably reduced the bias, a genuine incentive to be accurate, rather than a generic instruction to imagine harder, which the wider research doesn’t support as a fix on its own.

Exercise 2: The direct ask

Overview: Given that imagining someone’s perspective alone shows no reliable accuracy benefit, while having genuine information about them does, this exercise simply replaces the imagining step with actually finding out, in situations where that’s genuinely possible (Eyal, Steffel and Epley, 2018).

Steps:

  1. Notice a moment you’re about to guess how someone is reacting rather than finding out.
  2. Ask yourself honestly whether you could simply ask them instead.
  3. If you can, ask a specific, low-pressure question, “how did that land for you,” rather than a vague one, “are you okay.”
  4. Listen to the actual answer without immediately checking it against your original guess.
  5. Notice, afterwards, how close your guess actually was, since this builds a track record you can learn from over time.

Variations: In situations where asking directly isn’t practical, in the moment, asking afterwards is still far better than never checking at all.

Why it matters: This is the single highest leverage change available, given how large the accuracy gap between imagining and genuinely finding out actually is in the research.

Exercise 3: The confidence check

Overview: Because confidence in a read and the accuracy of that read are only weakly related, this exercise deliberately separates the two, rather than treating a confident read as a correct one (Ickes, 1997).

Steps:

  1. Form your read of how someone is reacting, as you normally would.
  2. Before acting on it, rate your confidence in that read honestly, on a simple scale from unsure to certain.
  3. Separately, ask what evidence the read is actually based on, their words, their tone, or your own assumption about how people generally react.
  4. Notice when confidence is high but the actual evidence is thin, that’s the combination most worth double checking.
  5. Over several instances, track whether your confidence and your actual accuracy, once you find out, are as closely linked as they feel in the moment.

Variations: This works well as a quick private habit, rating confidence silently, rather than something you need to say out loud.

Why it matters: Confidence is a poor guide to accuracy, which means the reads that feel most obviously correct are exactly the ones worth the most scrutiny, not the least.

Exercise 4: The specific difference

Overview: Under-adjustment happens partly because it’s easier to stay close to your own perspective than to genuinely work out how someone else’s differs. This exercise makes that work explicit rather than optional.

Steps:

  1. Choose someone whose reaction you regularly need to read, a colleague, a direct report, a stakeholder.
  2. List two or three concrete ways their situation differs from yours, their incentives, their history, what they’re accountable for.
  3. Before your next interaction with them, deliberately factor those differences into how you expect them to react.
  4. Afterwards, check how accurate that adjusted expectation actually was.
  5. Refine the list of differences over time, as you learn which ones actually matter for that specific person.

Variations: This is particularly useful for someone you consistently misread, since the same one or two differences are often doing most of the work each time.

Why it matters: This turns a vague instruction to empathise more into a specific, repeatable habit of naming what’s actually different about the other person’s position.

Exercise 5: The prediction check

Overview: The way researchers actually measure empathic accuracy is by comparing a prediction against what someone genuinely reports feeling afterwards, not by trusting the prediction on its own. Borrowing that method directly turns a one-off guess into an ongoing, checkable habit (Ickes, 1997).

Steps:

  1. Choose one relationship where accurate reading genuinely matters and pick it as your practice ground.
  2. Before each interaction, briefly predict how the person is likely to react to what you’re about to say or do.
  3. Afterwards, find out what actually happened, ideally by asking rather than by inference, and compare it against your prediction.
  4. Note specifically where the prediction was right and where it was wrong, rather than just whether it felt close enough.
  5. Carry what you learn into the next interaction, treating this as an ongoing habit rather than a one-off check.

Variations: Keeping a short running note of predictions and outcomes makes the pattern far easier to see than relying on memory alone.

Why it matters: This is what turns the other four exercises from occasional corrections into something you can actually track over time, checking your read against reality rather than trusting how confident it feels.

Exercise wrap up

These five exercises work together in a specific order: exercise 1 creates a genuine incentive to be accurate before you even form a read, exercise 2 replaces guessing with genuinely finding out wherever that’s possible, exercise 3 stops confidence from masquerading as accuracy, exercise 4 does the harder work of naming real differences when finding out directly isn’t an option, and exercise 5 turns all of it into a tracked, checkable habit rather than a one-off fix. If you only take one thing from this article, take exercise 2, replacing imagination with real information is the single change the research most strongly supports. Not every moment allows for that, though, which is where the micro-practices below come in.

Five micro-practices to start today

Five things you can start doing today, no dedicated time required.

  • Before assuming how someone feels about something, ask yourself honestly whether you could just ask them instead, then do it.
  • When a read of someone’s reaction feels obviously correct, treat that confidence as a reason to check it, not a reason to skip checking.
  • Notice when your read of someone else is really just your own likely reaction relabelled, then name one real way their situation differs from yours.
  • After a conversation that mattered, briefly compare what you predicted the person would think with what they actually said.
  • With someone you regularly misread, write down the one or two differences between you that usually explain the gap.

Conclusion

Reading people accurately isn’t about being a naturally empathetic person, it’s about correcting for a specific, well documented habit, anchoring on your own view and under-adjusting away from it. The single most reliable fix is also the simplest one available, asking rather than imagining, and the rest of what actually helps is less about sensitivity and more about treating your own confidence with a healthy amount of suspicion.

Reflective questions

  • Think of a recent read of someone’s reaction you felt confident about. How much of that confidence was actually based on evidence, rather than on how plausible your own reasoning felt?
  • Is there someone whose reactions you regularly misread? What’s the one difference between your position and theirs that you consistently underweight?
  • What’s stopping you from simply asking, in the situations where asking is actually available to you?

Explore further

Sources

Epley, N., Keysar, B., Van Boven, L. and Gilovich, T., 2004. Perspective taking as egocentric anchoring and adjustment. Journal of Personality and Social Psychology, 87(3), pp.327-339.

Eyal, T., Steffel, M. and Epley, N., 2018. Perspective mistaking: Accurately understanding the mind of another requires getting perspective, not taking perspective. Journal of Personality and Social Psychology, 114(4), pp.547-571.

Ickes, W. ed., 1997. Empathic accuracy. New York: Guilford Press.