What the data actually shows
We mispredict our own future feelings. Research on affective forecasting by Wilson and Gilbert shows that people are systematically poor at anticipating how events will make them feel and, especially, how long the feeling will last — a tendency they call the impact bias. We overestimate the emotional impact of both good and bad events, which means the things we expect to make us happy often deliver less than imagined, and we chase the wrong targets as a result.
Knowing the right thing to do is a weak predictor of doing it. Research on the intention–behaviour gap, including work by Sheeran, finds that good intentions account for far less of our actual behaviour than people assume — across health and other domains, a large share of people who genuinely intend to act never follow through. Intending to exercise, call a friend, or sleep more is only loosely connected to whether it happens.
Immediate rewards beat delayed ones in the moment. The activities most associated with wellbeing — social connection, physical activity, absorbing effort — generally require some up-front cost and pay off later, whereas convenient options like scrolling deliver a small reward instantly and demand nothing. Given a default pull toward what is rewarding right now, the easy option tends to win the moment-to-moment contest even when we know it serves us less.
Why this feels different from how it actually is
It feels like a personal failing — laziness, lack of discipline — because we experience the gap from the inside, as 'I knew what to do and didn't do it.' But the research reframes it as a structural mismatch, not a willpower deficit: prediction errors, weak intention-to-action links, and immediate-reward bias are features of normal human cognition, present in basically everyone, not symptoms of a uniquely flawed character.
It also feels confusing because we genuinely do know what helps. The knowledge is real; the problem is that knowledge is one of the weakest levers on behaviour. We tend to assume that understanding the right thing should be enough to do it, so when it isn't, we conclude something is wrong with us — rather than recognising that the knowing-doing gap is exactly what the evidence predicts.
And adaptation hides the evidence that the good habits work. Because we get used to the benefits of exercise, connection, or good sleep, their payoff becomes part of our baseline and stops being noticeable, while the cost of doing them stays visible up front. The very mechanism that should reward the habit quietly erases the felt reward, which makes the effortless alternative look comparatively better than it is.
What the research says to do about it
Reduce the up-front cost rather than relying on motivation. Since the wellbeing-supporting options lose the in-the-moment contest because they take effort, the most reliable moves make them easier and the low-value options harder: laying out gym clothes, scheduling the walk with a friend so it's a commitment, keeping the phone in another room. Shrinking the friction tends to do more than resolving to try harder.
Close the intention–behaviour gap with structure, not just resolve. The research on this gap finds that concrete plans specifying when, where, and how you'll act — sometimes called implementation intentions — meaningfully improve follow-through compared with a general intention. 'I'll exercise more' is weak; 'I'll walk for twenty minutes after lunch on Monday, Wednesday, and Friday' converts to action far more often.
Account for misforecasting by trusting tested patterns over predicted feelings. Because we mispredict what will make us happy, it helps to lean on what reliably works for most people — connection, movement, meaningful activity — rather than on our in-the-moment sense of what we feel like doing, which is biased toward immediate comfort. The activities you'll be glad you did are often not the ones you predict you'll enjoy most.
What the research says does not help
Simply learning more about what makes people happy does not close the gap, because knowledge is a weak driver of behaviour. Reading another article on the benefits of exercise or connection adds information you probably already have; the bottleneck is translating intention into action, which more facts don't address.
Relying on willpower and motivation alone is unreliable, since the immediate-reward pull operates in exactly the moments when motivation is lowest. Plans that depend on feeling like doing the right thing tend to fail precisely when you need them, which is why reducing friction and pre-committing outperform 'trying harder.'
Harsh self-blame for the knowing-doing gap is both inaccurate and counterproductive. The gap is a normal, near-universal feature of how prediction and reward work, not evidence of a weak character, and treating it as a moral failure tends to add discouragement without improving follow-through. Designing the situation works better than berating the person.
Real numbers in context
There isn't a single clean statistic for the knowing-doing gap, and any precise 'we follow through X% of the time' figure should be treated cautiously, but the direction is well established. Research on the intention–behaviour gap (including Sheeran's work) consistently finds that intentions explain far less of actual behaviour than people assume — a large share of those who genuinely intend to act never do — which is why concrete plans outperform good intentions.
The other two pieces are robust as patterns rather than numbers. Affective forecasting research (Wilson and Gilbert) shows we systematically overestimate how much and how long things will affect our feelings, so we aim at the wrong targets. And the basic asymmetry holds: connection and exercise cost effort now and pay off later, while scrolling pays off instantly and costs nothing in the moment — so the easy option tends to win unless you change the setup. None of this is a personal flaw; it's the normal shape of human motivation.