Do you really know what job will make you happy?

Why are so many people dissatisfied with their jobs? A big part of the problem is that we’re pretty bad at predicting how happy things will make us, or how long that happiness will last. We think, for example, that winning the lottery will make us much happier in the long run – but it probably won’t.(1) This has some serious ramifications for career choice.

In an earlier post, I concluded that there’s a much closer connection between careers which make us happy and careers that make a difference than we might think. One side of this relationship is that we’re more likely to make a difference in a career we enjoy. This means that even for the effective altruist, factoring in our own happiness at least to some extent when making career decisions is important. However, what this doesn’t mean is that we should simply follow our intuition (or our passions) about what makes us happy, and then factor this in to our decisions.

We have a tendency to overestimate both the positive impact of a “good” event (such as winning the lottery), and the negative impact of something “bad” happening (such as breaking a leg.) One study showed that accident victims often turn out to be happier a year after their accident than they expect to be.(1) Another asked professors to predict the long-term impact that either receiving or failing to receive tenure would have on their happiness: and lo and behold, in both cases the emotional impact (whether positive or negative) was overestimated.(2) If we can’t even predict what kind of sandwich we will want for lunch next week (3), then predicting what kind of career we will enjoy in five or ten year’s time seems like a hopeless task.

Why are we so bad at predicting what will make us happy? When we try to predict our emotional reaction to future events, we tend to simulate: we imagine ourselves in a future situation, which elicits an emotional reaction in the present. We then use this emotional reaction as a predictor of how we would react in the given situation.(4) So when I imagine feeling anxious sitting an exam, what I really do is imagine sitting an exam, which makes me feel anxious, and then take this anxiety as an indicator of the feeling I would expect to have when I actually sit the exam.(5)

The way in which we use simulation to predict our future emotional states provides a route into understanding why such predictions often go so far awry.

Our current emotional state affects our future predictions

Since we predict our future emotions based on an emotional reaction in the present to a simulated situation, our present emotions are liable to affect our predictions. A tendency to project our current preferences onto a future event is known as the projection bias. People at a gym who had just worked out on a treadmill (and were therefore thirsty) were much more likely to predict that thirst would be more unpleasant than hunger if lost in the woods, than those asked before using the treadmill (who were not thirsty).(6) It seems that when trying to imagine how they would feel when lost in the woods, the thirsty people were unable to ignore their current thirst in judging their future reactions.

Making big career decisions can often require us to predict how much we will enjoy a job in several years’ time, and it seems likely that by then some of our tastes and feelings will have changed – our desire to be working in the busiest part of London may be less strong aged 30 when we want to settle down with a family than aged 25, for example. The way in which we use simulations to predict future happiness suggests that we find it difficult to account for changes in our current emotional states when predicting how we’ll feel in the future.

We rely on the past to predict the future

In simulating future events, a big part is played by our memories of similar past events: we base our judgement of what a doctor’s appointment will be like largely on our memory of past appointments, for example. However, our memories are often unrepresentative of the past events themselves, and therefore also of future ones.

Unusual and recent events tend to be more memorable, meaning they play a greater role in shaping our judgement of future events. When asked to remember a time when they missed a train, people tend to recall the most negative such experiences.(7) This makes them likely to overestimate how painful their next train-missing experience will be.

Similarly, we might overestimate how much we will enjoy (or dislike) a certain component of a job if our simulation of it is skewed by the more extreme of our past memories.

Simulations only account for essential features

According to construal level theory, we tend to conceptualize future events much more abstractly than immediate ones. A similar point is that our simulations are essentialist – they are comprised of only the most important features.

This makes sense: simulating future events is often hard to do, and we can’t possibly focus on every minute detail, so we just include the most important ones. The problem is that our perception of what are the most important features is fallible – leaving our simulations susceptible to the focalism bias.

By basing our emotional predictions on what we deem the most important features of a future event, we may miss out on some non-essential features which would regardless make a big difference to our emotional responses.

One explanation for why the tenured professors ended up less happy than predicted might be that they focused overly on certain positive features of their tenure: the sense of achievement and recognition, at the neglect of others: increased working hours and dull departmental meetings, for example.

When deciding which career to pursue, therefore, we are likely to focus our attention on factors which come to mind easily such as salary and working hours. This might lead us to ignore other factors which are actually crucial to predicting happiness like mental challenge.

Initiatives like 80,000 hours help to counteract the focalism bias in career choice, by drawing attention to factors which people don’t tend to focus on.

We forget that emotions fade over time

A final point is that when predicting how a future event will make us feel, we tend to focus on its early features at the neglect of later ones. This might help explain why the lottery winners and accident victims overestimated the impact of the event: they imagined how they would feel immediately after winning, or having the accident, and assumed that this feeling would last.

However, emotions fade in intensity over time, and we tend to underestimate this and the speed at which it happens.(8) This is the durability bias: a tendency to overestimate the emotional impact of an event in duration.

Susceptibility to the durability bias means, for example, that we might expect a job with immediate rewards (like a high starting bonus) to make us much happier than it actually will in the long run.


So although we need to factor in our own happiness when making career decisions, simply using our intuitions to predict what makes us happy isn’t the way to do it. We’re overly biased by our current preferences, rely too heavily on selective memories, miss out potentially important features and underestimate the extent to which our emotions fade over time.

Taking account of what we think will make us happy is pointless if these predictions aren’t remotely accurate – especially if we’re weighing this up against other important factors. In the next post I’ll talk about how we might try to factor in our own happiness whilst avoiding these biases.

(1) Brickman, P., Coates, D., Janoff-Bulman, R., (1978) “Lottery Winners and Accident Victims: Is Happiness Relative?”, Journal of Personality and Social Psychology 36, 917-927

(2) Blumberg, S., Gilbert, D., Pinel, E., Wilson, T., (1998) “Immune Neglect: A Source of Durability Bias in Affective Forecasting”, Journal of Personality and Social Psychology 75, 617-638

(3) A simple experiment provides two groups of participants with free sandwiches for a week – the first group gets to pick which sandwich they want each day, whereas the second have to choose their sandwiches for the week in advance. The people in the first group tend to choose the same sandwich each day, and are reasonably happy with their choice. The second group, however, who have to predict what they will want in advance, tend to opt for variety. This group are significantly less happy with their choices than the first – it seems they don’t like variety as much as they thought they would. See Gilbert, D. T., & Wilson, T. D. (2000) “Miswanting: some problems in the forecasting of future affective states” in Feeling and Thinking: The Role of Affect in Social Cognition, 178–197.

(4) Gilbert, D., Wilson, T. (2007) “Prospection: Experiencing the Future.” Science, 317, 1351-1354

(5) Work in neuroimaging suggests that much of this simulation goes on in the pre-frontal cortex. Interestingly, patients with damage to this area seem unable to simulate future events, having severe difficulty answering questions such as “What will you be doing tomorrow?” See the above paper (4) and Dan Gilbert’s “Stumbling Upon Happiness” for more details on this.

(6) 92% of the “after treadmill” group predicted finding thirst more unpleasant, compared to 61% of the “before treadmill” group: Van Boven, L., Loewenstein, G. (2003) “Social Projection of Transient Drive States”, Personality and Social Psychology Bulletin 29, 1159-1168

(7) C. K. Morewedge, D. T. Gilbert, T. D. Wilson (2005), “The least likely of times: how remembering the past biases forecasts of the future” Psychological Science 16, 626-630

(8) Finkenauer, C., Gallucci, W., Pollmann, M. (2007) “Investigating the Role of Time in Affective Forecasting: Temporal Influences on Forecasting Accuracy” Personality and Social Psychology Bulletin 33, 8