Would We Make Better Decisions If Life Had Instant Feedback?

Imagine buying a product with a photograph of the disease it may eventually give you printed on the package—and buying it anyway.

That is a strange transaction.

The warning has not been hidden in paragraph 47 of the terms and conditions. It is right there, interrupting the design. A person can see it, understand it, believe it and still smoke the cigarette.

This is often treated as a failure of information:

If people really understood the danger, surely they would stop.

But many do understand.

The information may have successfully travelled from the package to memory and from memory to belief. The failure appears somewhere closer to the next decision.

At that moment, one side contains a cigarette, a craving, a ritual, perhaps friends, and—for a person who is dependent on nicotine—relief from withdrawal.

The other side contains an increased probability of illness in a future that may be decades away.

One is an experience.

The other is a forecast.

The wrong lesson arrives first

Life already gives us instant feedback. Unfortunately, the immediate feedback often rewards the behaviour whose real cost arrives later.

Smoking can produce stimulation, ritual or relief now. Much of the health cost arrives later.

Skipping exercise produces rest now. Fitness arrives later.

Spending produces enjoyment now. Financial security arrives later.

Procrastination might be the cleanest example because it does not require a clinical addiction or a forty-year follow-up study.

Suppose a report is due on Friday.

On Monday you open the document. The empty page produces discomfort. You close it and watch something else.

Relief arrives immediately.

An excellent result, according to the fastest available data.

The panic arrives on Thursday night. By then, Monday’s small act of avoidance is one event among hundreds. The behaviour was rewarded at once and punished three days later.

Your brain has learned something true but unhelpful:

Closing the document makes this unpleasant feeling disappear.

The world is a teacher with dreadful timing.

The future arrives as a summary

Behavioural researchers use delay discounting to describe how an outcome’s subjective value tends to fall as it moves farther into the future. A larger-later reward can lose to a smaller-sooner one, even when waiting would leave us better off.

That does not mean people consciously decide that old age is irrelevant. It means distance changes the force an outcome has in the present choice.

There is even evidence of a sign effect: in some experiments, delayed gains were discounted more steeply than delayed losses. (Thaler, 1981) But that wrinkle is not required to explain the basic problem. Exercise offers a delayed gain wrapped in an immediate cost. Smoking offers an immediate experience whose largest costs are delayed and uncertain. Both make the present unusually persuasive.

In one study, current smokers discounted delayed hypothetical money more steeply than never-smokers and ex-smokers did. The researchers were careful about what this comparison could not establish: smoking might increase discounting, people who discount the future more might be more likely to smoke, or both processes might interact. (Bickel, Odum & Madden, 1999)

Smoking is also more than a lesson about time. Nicotine dependence, withdrawal, habit and environmental cues matter. “People discount the future” is an incomplete account of addiction.

Still, the comparison exposes something important.

At the moment of choice, a cigarette is concrete. Disease is statistical. The craving belongs to the present self; the warning describes someone the present self has never met.

The future may carry enormous objective stakes while exerting very little psychological force now.

Importance and felt importance are not the same thing.

Knowing is not the same as receiving feedback

We often collapse four different events:

• knowing that a consequence exists;
• feeling that consequence as relevant now;
• receiving a signal after an action;
• receiving a signal that accurately represents what matters.

A warning can provide knowledge before the action. A craving can provide felt experience during it. Relief can provide immediate feedback after it. Illness may provide much more accurate feedback decades later—when thousands of possible causes and choices sit between action and outcome.

This is why more information does not always repair behaviour. The sentence “smoking increases your risk of disease” can be understood perfectly while losing the competition for attention at the exact moment it matters.

In machine learning, a related problem is called credit assignment. If an action is followed immediately by a reward or penalty, it is relatively easy for a learning system to connect the two. Touch a hot stove; feel pain; do not repeat.

If the consequence arrives after a long chain of actions, identifying which earlier choice deserves credit or blame becomes harder.

Smoke at nineteen; develop heart disease at sixty-one.

Skip one workout; observe almost nothing.

Delay saving for one month; remain financially comfortable.

Humans are not simple reinforcement-learning agents. We can use language, imagine counterfactuals, learn from other people and commit ourselves to rules. That is precisely why a picture on a package can matter at all.

But reasoning must still compete with a nervous system receiving vivid, immediate signals.

We may be trying to learn from a world with terrible reward timing.

What if life scored us instantly?

Imagine that consequences stopped being shy.

Every cigarette immediately reduced a visible health score.

Every skipped workout displayed the fitness it would eventually cost.

Every cruel comment produced, on the spot, the relationship damage it would create three years later.

Every unnecessary purchase flashed:

Estimated retirement delay: 17 minutes.

Surely we would make better decisions.

Perhaps we would exercise more, save earlier, apologize faster and complete the report on Monday like the deeply suspicious people who complete reports on Monday.

But what exactly would the score measure?

If exercise produces immediate fitness points, some people will optimize the points until recovery looks like laziness.

If parenting produces an immediate child-happiness score, parents may remove every frustration and raise a child who has never practised recovering from disappointment. That is the same objective-function problem hiding inside the question of whether parenting is an optimization problem.

If work produces a live productivity score, people will prefer countable tasks to uncertain exploration. The score may improve while the valuable work disappears—a faster version of what happens when a measurement becomes the target.

And social media already gives us immediate feedback: likes, views, replies, streaks and the tiny red circle that apparently contains an emergency. Those signals are fast and precise. Whether they represent the life we want is another question. (Is Social Media Training Us—or Are We Training It?)

Faster feedback improves learning only when the feedback points toward the thing we actually care about.

Otherwise, we learn the wrong lesson sooner.

Would it make us better people?

There is a moral version of the thought experiment.

Suppose lying caused immediate physical pain. Suppose cruelty made us instantly feel the injury experienced by the other person.

We would probably lie and harm less.

Would we be more honest and compassionate, or merely more constrained?

This resembles the problem in asking whether we can know our humility before we have power. Behaviour under a constraint is evidence of character, but it is not the same test as behaviour when the constraint disappears.

A person who tells the truth only because a morality dashboard subtracts twelve points may be well behaved. I am less certain they are honest.

Instant consequences could improve conduct while leaving character underexamined.

They could also make us easy to control. Whoever designs the feedback gets to decide which behaviour feels good now, which feels costly and which remains invisible.

The problem is no longer simply delayed feedback.

It is governance.

The future may not need to arrive early

Perhaps we do not need a perfect dashboard. Perhaps the future only needs to become less abstract at the moment of choice.

In experiments on episodic future thinking, participants imagine specific personal events that could occur later. In one study, presenting people with cues to their own future events reduced their preference for smaller-sooner rewards; the more vividly they imagined those events, the more future-minded their choices became. (Peters & Büchel, 2010)

A later randomized experiment with smokers compared imagining personal future events with imagining recent events. The future-thinking group discounted delayed rewards less and earned fewer cigarette puffs in a laboratory self-administration task. (Stein et al., 2016)

These are not demonstrations that imagining your sixty-year-old self cures addiction. A laboratory choice and a smoking task are not a lifetime, and long-term behaviour must survive cravings, routines and environments.

But they suggest an intriguing possibility: the future does not always have to arrive physically before it can influence the present. Sometimes it can be made more detailed, personal and available to attention.

That may be part of what deadlines, automatic savings, public commitments and intermediate milestones do. They translate one distant outcome into nearer signals.

The danger is that the translation becomes the target. The milestone replaces the mission. The daily score becomes more real than the life it was supposed to represent.

My current hunch

Perhaps people do not ignore the future simply because they misunderstand it.

The present arrives in high resolution. It has sensation, emotion, effort and relief. The future arrives compressed into a sentence: healthier later, poorer later, regret later.

Meanwhile, the action that creates the later problem may be teaching the opposite lesson now.

Avoidance works.

Spending feels good.

The cigarette removes the craving.

So the aim should not be instant feedback at any cost. A rapid but inaccurate signal can make us reactive, manipulable and extraordinarily efficient at optimizing the wrong thing.

The better aim may be faster contact with the real consequence: making a future state vivid enough to enter today’s decision without pretending that health, character or a relationship can be reduced to one number.

I am still left with a difficult question.

If we could make our future selves feel psychologically present—without inventing a score that distorts them—how many decisions would change?

And when the present self and future self still disagree, which one should get the final vote?

Until the next strange question,

Osagie