Sunday, 18 September 2011

Intrinsic motivation - I doubt the research 1/2

Deb Louison Lavoy shared a TED talk with me, Dan Pink on the surprising science of motivation

Briefly summarising that, Dan shows proof of extrinsic motivation (pure money in this case) actually having little to none and sometimes even a negative effect on people's performance. In short, when the reward for a task was at its highest, people did actually perform worse than when it was a medium reward: in the experiment conducted there were three kinds of rewards: low, medium and high

I want to write two posts on this. This one is about me hairsplitting the very research in question that underlies Dan's passionate and entertaining TED talk. The second one will be about my opinion on intrinsic (and of course, extrinsic as well) motivation

Update 19th September 15:05 CET: I exchanged some tweets yesterday with some people about this post, and the outcome baffled me: people attacked me about intrinsic motivation in general, pointing me to research even (how ironic), and I even got called jerk and idiot.
Please read the above carefully before you react. I also changed this post and added one line below in which I clearly state that good performance in this test would entitle a participant to the equivalent of one month's worth of salary - for half a day's work. Performing very well would "only" double that. Would any performance below good be rewarded? No, not at all. So only good or very good performance would be rewarded. Well, I think it suffices to say that merely performing good - according to the research team - would be considered to be outrageously well rewarded by any participant. I'd sure as hell make sure that I perform just good, so there's hardly any incentive to perform very well
Hence why I consider this very research to be fraudulent and urge it to be completely ignored from a scientific point of view; and that's the only focal point of this post

The research paper can be found here - I'm referring to Experiment I only

What is it about? In 2002, 87 rural Indians took part in it. Half of them owned a TV, half of them owned a bicycle, 7% owned a telephone, none owned a car, a quarter was female and a quarter had no education at all. Average per person annual income for all? 6,000 rupees - remember that.
Their ages weren't mentioned at all, which I think is a huge omission. As the study says, "The experiment was conducted by local research assistants from Narayanan College at Madurai, India, who were naive to the hypotheses"

Each of them participated in the experiment once, randomly assigned either a low, medium or high reward. The experiment consisted of 6 games, in which three performance levels could be reached each time: none, good and very good. For good performance, reward would be 50%, for very good, 100%. Rewards for each game were 4 rupees for low, 40 for medium and 400 for high. My guess of each game's duration is that they all took 5 minutes on average. Count in some overhead for explanation, the entire experiment would take 1 hour per person, well let's just make that 2 then, no, half a day to rule out everything else and be on the very safe side, OK?

Now, let's make that intelligible for an average US citizen. Taking into account education, in 2003 a day's work would earn someone $22,718 (High School drop-out) - I think that's more than a fair comparison for these people who on average had 5-6 years of education, a quarter of all having none whatsoever. Given the annual income of 6,000 rupees, the six games, the 4, 40 and 400 rupees reward, and reward percentages being either 0%, 50% and 100%, the rewards in rupees and dollars is shown above

Explanation: the annual income is 6,000 rupees or 24,000 dollars (just upped that from $22,718 so it's easily multiplied by 4).
With 52 weekends and 4 weeks of holiday, there are 240 working days in a year. Average daily income would thus be $100, for half a day (we just suggestimated this test to take four hours, didn't we?) that would be 50 dollars.
If you look at the maximum rewards (6 games times the rewards), one person could win 9,600 dollar if he'd achieve very good performance on every test. That is the equivalent of 96 days of work, almost 5 months. If you take into account the half day of "inspiration", that means that someone would get rewarded 200 times as much as usual - now I don't call that a high reward, that's humongously obscenely HIGH

In my world, the difference between low, medium and high income is a factor 5. A low income in the US would be $20,000 annually, medium would be $100,000 and high would be half a million dollars per year. So that $9,600 would be what a millionaire makes in a week - and it would take you only half a day: that's just dead wrong. The high reward for a half-day task should be what a millionaire makes in half a day: 1,000 dollar. Medium should be 200 dollar, and low 40 dollar: because that's what someone with a $20,000 annual income makes on half a day's work: $40

So what if a person would average out on it all? Medium reward and medium performance? Then he'd receive 480 dollars - that's the equivalent of an entire workweek! In just half a day of bumming out an some experiment! Not bad hey?

What if you scored the absolute lowest reward, besides nothing, and were "incentivised" with a low reward every single time?
In the very unlikely likelihood that you'd get low incentives on all 6 games, and achieved medium results, you'd still receive $48: exactly what you'd make on an ordinary work assignment.
On 6 experiments, that's a very low chance, because randomly being assigned incentives means that you get a low one twice, a medium one twice, and a high one twice: that means if you don't really do your best at all, but just get a good performance in every game, you'd get $1,776 - that would be 42 times as much as you'd make in 4 normal working hours

Here's where the experiment is clearly fraudulent. I don't even consider this an omission or carelessness, this can't be anything else than an attempt to falsify the outcome of it, rig the experiment, game the system: you name it.
Whatever is called low reward here, is well over the average medium reward for those who participated in the experiment (taking into account the 4 hours for the experiment in stead of 1 or 2, the upping of $22,817 to $24,000 and the fact that I assume the participants to not sacrifice any work over this experiment, or having been compensated for it - but feel free to ignore that last remark)

But, how does that explain the high rewards scoring lower than medium or low? It doesn't - but I can

Once I was fluent in German. My German was so perfect that people asked me "Where are you born, North Rhine-Westphalia or Rhineland Palatinate?". That was from the age of five until early twenties.Meanwhile, I did High School exams, one of which was German. Of course I had a straight A for German, that had been the case since I took up German 5 years before that.
The exam was tough. Extremely tough. In fact, I finished as one of the last people in my class - imagine that. Afterwards my teacher came up to me, clearly worried, and I complained about the exam being so hard. He stared at me "Are you kidding? This was the easiest exam ever! There are going to be repercussions and discussions about this one, take my word on it - this was unfairly easy, honestly"

I was shocked. Yes it had all seemed uncanningly easy, but I assumed that every one of those was a trick question. I had deliberately chosen the difficult answers over the straightforward ones. To make a long and awful story short: I got a D on the exam, and as that was 50% of my endgrade, I ended up with a B- for German - which should have been an A+

So, can relatively easy tasks combined with an obscenely high reward lead to decreased performance? Yes - for me it certainly did

0 reacties:

Post a Comment

Thank you for sharing your thoughts! Copy your comment before signing in...