hypios

Dysfunctional Incentives?

Former Al Gore speechwriter turned business-world bestseller, Dan Pink recently gave an excellent 19 minutes talk at TED2009.   (For those of you not familiar with TED, hasten to their website—it offers an extensive video archive of important contemporary thinkers on technology, design and innovation).

In this talk, Pink tells us that most of our ideas about what motivates performance are wrong—or at least not well-adapted to important tasks of our times. His main point: our beliefs about incentives come from attempts to motivate tasks that we no longer do.

The mechanical, rule-following tasks of the 20th century have been replaced by those requiring open-ended, creative thinking.  For creative tasks, contingent incentives (if you do X, you will get Y), have a negative impact on performance.

Correlatively, studies show a negative effect with the addition of a competitive element to a “task environment”.  A contingent-reward structure often involves setting a solver against his or her peers performing similiar tasks (only some of them will get Y for doing X).

Pink highlights two studies which clarify the interdependence of contingent reward and competition.  In one experiment, participants are asked to complete a challenging cognitive task, i.e. to solve it, participants need to operate what can be called a change of a perspective, (“the candle problem”). They are divided into two groups.  One group is told that their work will be used to set a “norm” or performance median.  The other group is told that they will be timed and rewarded according to their performance (relative to their peers).  Not only did the norm-setting group do better at their task than the pay-for-performance group, but the more that was on the line (in terms of reward), the worse participants performed the task.  When the task was slightly changed, implying less creativity and more mechanical behavior (what Pink calls “the candle problem for dummies”) contingent incentives do work.  Study after study appear to corroborate this evidence.

Pink does not spend too much time speculating on why contingent rewards would have this effect.  Instead he uses his time to discuss the lessons we ought to learn from this research: pay incentives don’t work when tasks are cognitively challenging;  people do better in less competitive, more autonomous environments. So whereas payment is the best incentive for mechanical tasks like those performed on mturk, payment alone would not seem the best motivation for more complex tasks like those performed on hypios. There has to be some intrinsic motivation: Solvers have to  feel like solving problems “for their own sake” is rewarding.

I think Pink is largely right.  In fact, the work environment he describes, ROWE (results-only work environment)—which allows for relative work autonomy—is what we’ve adopted here at hypios.  What I’d like to do here is to speculate a little about why solvers might have reacted in the way that they did in the experiments and try to show that we shouldn’t be too quick to draw conclusions about the efficacy of either financial incentives or competition in general.

The Creativity Jam

Pink focuses on what he calls contingent incentives.  Studies that Pink cites show that an incentive tied directly to the outcome of a single (timed) task cause worse performance for cognitive tasks.

Studies have also found that setting specific and difficult goals significantly increases solver performance even when the goals are out of reach.  These same studies have found that, in the context of these specific and difficult goals, adding competition hurts performance when competition comes in the form of measurement against peers in view of reward.  Correlatively, studies have shown that cooperative environments were superior to competive ones for enhancing task performance.

So it appears that competition and contingent reward work together to jam creative thought, at least under conditions where the reward was directly tied to the performance of a specific task.  You might say that together they seem to create a virtual axe over your head, which is counter productive to creative results. Working in an autonomous environment where you set your own specific goals is most conducive to them.  What none of these studies have shown (and they acknowledge this) is what happens when competition and incentives are on the horizon of an activity rather than directly tied to it.

Challenging Pink:  Not all competition is zero-sum

If Pink is right, why do prize competitions—competitions that involve pay for winning performance—like the Netflix prize seem to work so well as creative-solution generators?  Does offering payment for solution (as hypios does) discourage or “de-motivate” solvers?  Can’t a little friendly competition (against a rival institution, say) be an added driver (even if it may not be as important as collaboration with peers)?

Pink is surely right when he says that intrinsic motivation (say, finding a task interesting) is more important than most anything else when it comes to the performance of high-level cognitive tasks.  But it is hard to imagine an intrinsically motivating task withdrawn from a horizon of competition and reward.  What makes it worth doing is exactly what makes it valuable to others.  There will always be rewards (in the form of glory and gold) and competition for any genuinely worthwhile task.  And people often (though by no means exclusively) get clues about what tasks are worthwhile from the array of incentives that surround a given task-environment.  How can we put these sorts of intuitions together with the finding that Pink highlights in his talk?

Just as in economic theory, the price of something is thought of as a signal sending a message to a consumer or producer: financial incentives send messages that something has value for a seeker and even for a society as a whole.  As such they focus attention on certain types of problems, issues, tasks, marking certian ones as valuable.  Indeed, the X Prize foundation, which runs prize competitions with purses exceeding 1o million USD, uses these large sums to counteract what they see as a market failure, to signal attention to a series of goals that they see as undervalorized (by the market) but that they have identified as key for social advanacement.  If there is only one winner, this does not seem to discourage pariticpants nor limit personal investment in pursuit of the prize.

The Ansari X prize, promoting civilian space travel, had dozens of participants investing millions of their own funds in pursuit of the goal.  This is because the competition is not viewed as a zero-sum game.  Competitors gain from the competition in terms of early entry into a promising market, experience and technology exposure.  Here the award money structures the horizon, but is not linked to strategy development nor to the completion of individual tasks.  How a solver gets to the winning solution is set by that solver or team, and not prescribed by a carrot-wielding outsider with a stopwatch.

Now let’s think a little more about competition.  Is collaboration really the only game in town?  I would say (and for reasons that Pink is willing to call “ideological” and “lazy”) collaboration has traditionally gotten short shrift.  But, there’s evidence from studies of  organizational competition that show that organizations that are “neck to neck’” spur innovation.   In the Netflix competition we saw two teams—agglomerations of single players and smaller teams—that were, in the end, competing neck-and-neck.  Neither the competitive element nor the prize money ($1 million) seemed to create a dysfunctional task environment among the leaders.

Let me offer a  reason why.   Netflix created something called a leaderboard, which showed the performance of all pariticpants and thus allowed participants to gauge their performance against their peers in relative “real time.”  In the cited studies (and in many real “task environments”) participants are in the dark about others’ performance, causing them to feel a loss of control and autonomy with respect to their task.

In the Netflix case, the reward was tied to a specific and challenging goal (10% increase in the quality of the algorithm) and contest rules made it so that teams were given time to match or catch up with the winning team.  Basically, this mimics the information that rival firms would get about their competition in a more or less free market.  Now it’s important to not generalize (as I’ve been doing) too quickly between the response of  individuals and teams in a task environment.  This deserves much more study.

Pink points us towards some very important insights into the nature of motivation.  Financial incentives not only cannot replace intrinsic motivation, they can acutally inhibit it.  Intrinsic motivation, which comes from internal values and goals matching external problems and tasks, is decreased when factors of reward and punishment change the task environment and make the participant feel like they are no longer in control of the situation.  With the solver in control, goal, guts, glory and gold all factor into what challenges he or she will take up.  Watch Pink’s video (if you feel motivated to do so) and then let us know your thoughts. We promise, we won’t pay you for it

Share this post!