Tag: behavioural science

An Indian Survivorship Bias Example

Many of you must have seen that drawing of a World War II plane with red dots on it.

It explains Survivorship bias, a bias that statistician Abraham Wald figured out.

Pic: source

Simply put, survivor bias is our tendency to view a situation or pattern as a comprehensive representative sample, often without considering what might be missing from that picture.

For instance, the WWII plane with red dots was a study Wald and his team carried out to determine which parts of returning Allied jets were hit the most by enemy gunfire, so as to reinforce those parts and make them stronger.

It was then that Wald realized that those parts were actually stronger, as all those jets had made it back to base safely. So instead, they focused on reinforcing the other parts, since clearly jets hit on other weaker sections never survived to tell their side of the story.

In India, there has been a belief among many of the older folk that children or people with big/long ears live a longer life, compared to those with smaller ears.

Interestingly, while most of our body shrinks with age, our nose, earlobes and ear muscles keep growing. Which means our elders had the concept backwards.

It was not that those born with bigger ears lived longer. But rather that those who lived longer, had ears that simply had a longer time to keep growing, and are therefore, relatively bigger in size.

#behaviour #behavior #bias #behaviouralscience #behavioralscience


Sample Size of One: Towards a Possible Solution

This post explores an alternative to fix the replication crisis (particularly in the behavioural science and economics fields, and if relevant, in other fields too). This post is in continuation to an earlier post titled Sample Size of One: The Rose Negotiations. It would help to read that one first before coming to this one.
What can we do to solve our human desire to create or find patterns and thumb rules to how we function? Or to find keys to getting humans to behave in a particular manner, be it to drive a more healthy culture, or improve finance sense among populations. Especially when few patterns exist. And when our desire might be overly simplifying a pattern which might be far broader than what we might want it to be.
Here is a broad suggestion towards what a possible solution looked like, at least in my head.
Consider a “hypothetical” scenario where a group of researchers wants to find the effects of an ‘opt-in/opt-out response’ for organ donation.
Up until now, behavioural economists or scientists would identify a large, diverse study group of volunteers, and conduct the experiment. Let’s suppose at the end of the study, they found that 70% of respondents opt for the organ donation program when the form requires them to physically opt-out of organ donation.
Now, a non-profit across the world tries this tactic on a local population, but perhaps has a less than encouraging (and far less than a 70% success rate) outcome. This leads to questioning the research findings, and the broader hue and cry around the reproducibility and replicability of such studies/ experiments.
For a moment, consider currencies. They are always fluctuating, and there is a definite exchange rate to convert between any two currencies at at a given point in time.
Or consider diverse marketplaces across the world. Any single product would be differently priced in these different marketplaces. And within a single market, the price variance might not be too much. But it might vary if you went to a market in the next village.
Coming back to trying to find an alternative to traditional experiments that try to find thumb rules to then apply to social or business causes.
The alternative I see, is where the economist or scientist creates a simple experiment or study around the hypothesis they would like to test. They would then put it on an online platform (and share it with their colleagues and counterparts across the world, who would then deploy it among local populations).
The experiment would be introduced via a website. Deployment could be done online with voluntary participants, or random people. The experiments would run in perpetuity (hence online), and results of the same would keep evolving over time and geographies.
The outcome for the same opt-in/ opt-out hypothesis with this alternate deployment might look something like this:
The experiment is designed to be unbiased, simple (easy to deploy without the original team of researchers being physically present), and yet robust enough to provide meaningful data.
The results of this experiment would not be captured as a single value (like 70% in the first hypothetical scenario), but rather as a function of (age/sex/location/study response/point in time).
As a result, what the outcome might look like, is diverse data points from across the world at diverse points in time.
It is possible that patterns will emerge in localized groups, or even at a nation-level for some experiments (since respondents or the general population might share a similar national history, current political and economic environment, and similar fears and concerns – whether it is about inflation, unemployment, or a multitude of other variables that were possibly getting ignored when a research study focused on finding a thumb rule.
With a global, perpetual study, for the same opt-in/ opt-out experiment, we might perhaps get results like an average of about 65% in Mumbai, India, but a 20% on the outskirts of Mangalore, India, and maybe even an 80% in Itanagar, India.
That way, researchers and anyone trying to use these research findings would be mindful that it isn’t a one-size-fits-all finding. But rather that perhaps (cautiously), one might expect to get a similar response to an organ donation campaign in a town in Country 1 and a city in Country 2, because their outcome values over a particular period of time have been similar.
And these values that emerge across individuals and locations are not fixed values. They are ever-evolving, to reflect the evolution of humans in a particular society, given the context of its changing sociopolitical and socioeconomic landscape, among other variables. So perhaps the same individuals too could participate in the same experiment multiple times over the years, with different results. In that sense, it would be similar to taking an IQ test or an MBTI test.
Which means, the same non-profit that is driving an organ donation exercise in a particular country in a particular year, would refer to the current result outcomes for different parts of that country, to determine what strategies they might have to employ (government intervention, financial incentives, etc.), towards driving a more successful change effort.
An obvious extension of this proposed solution will be coming to a post soon.
In the meantime, it was still challenging conveying the problem and my solution concept to people. So I started working on a simpler way to do that, and here it is.
#SampleSizeOfOne #BehaviouralScience #BehaviouralEconomics

The Behaviour Triangle


A humorous take on the paradox that exists between the views or tendencies of us common humans, versus that of therapists, who seem to take the more empathetic approach, versus some behavioural science practitioners who try to leverage behavioural knowledge to grow business without it necessarily being beneficial to customers themselves.

A related interesting read someone shared: Nudge Theory needs to take more external factors into account.

#Humour #BehaviouralScience #BehaviouralEconomics #Nudge #psychologist #Behaviour #Behavior

Sample Size of One: The Rose Negotiations

Image: source

The Replication Crisis is an ongoing crisis where it has been difficult or impossible to reproduce findings of scientific studies.

The field of behavioural science too, has had its challenges with replicating past research findings. Some years ago, peer-reviewed scientific journal, Nature Human Behaviour, attempted to replicate 21 social and behavioural science studies published in the top peer-reviewed journals, Nature and Science. It could replicate only 13. Other such studies conducted too, resulted in disappointing results.
Is it surprising if behavioural science and behavioural economics research findings are difficult to replicate? Till recent decades, many studies were undertaken by professors on captive university students; a long shot from representing world diversity. Findings from one country could throw up different results in another country based on many variables like the history of that nation, recent and current economic progress, poverty levels, trust levels, level of ethics in government, enforcement and business, among other factors. We see diversity even in our interactions with foreigners on social media.
And yet, we as humans, are always trying to find a simple common denominator. A thumb rule. A recipe or formula that we would like to think would apply to the world population.
Instead, what if we looked at behavioural science endeavours the other way around? Why struggle to have larger and more diverse sample sizes to represent world diversity and improve study accuracy? Instead, what if we started with a simpler sample size? One that we are more sure of, and that offers more accurate data points. What if we start with us as individuals? The observations, feelings, rationale or reactions of the individual (hence ‘Sample Size of One’)? And from there, cautiously see if and to what extent, it applies to other individuals or groups?
Whenever we experience a situation, we could try and assign values to various parameters. Then, similar situations created for others in other parts of that city, country or the world, would give us more data sets. We could then look for a single line passing through diverse cultures, or spot similarities in diverse groups. Similarities in similar groups from contrasting countries too? Or not.
What we will have, is readings across these situations or experiments across countries and cities. At present, if an experiment with a sample size of 10,000, finds 70% respondents behave a certain way in a scenario; we extrapolate it and believe 70% of world population might behave in a similar way.
Instead, what if we had that scenario, but had different data sets for different locations? That way we might find clearer patterns (there goes the human in me again) among groups in diverse cultures where a certain improvement intervention might respond similarly to another one which had similar outcomes from an experiment.

This thought occurred to me during a recent festival and an interaction with a flower woman in the market. Here’s the story.

Sample Size of One: The Rose Negotiation

My family is a bit religious. During Dussehra puja a few weeks ago, I was back at the market, buying flowers and fruit.

A woman in the market has a flower stall and sells bouquets. I always buy roses from her. On regular days, a rose costs INR 10. On festive days, she sells them for either INR 12 or 15.
I asked her how much for one.

“INR 15”, she smiled and replied, “but you can have them for INR 12. How many do you want?”

I asked for ten. Like always, I asked her to cut them to a particular length. I said I’ll pick up the other stuff on my list and come back for the roses.

Back at her stall 15 minutes later, she said the total was INR 150.

I said, “but didn’t you say you’d give them at INR 12 a piece?”

Now she seemed confused, like she had goofed up the prices.

In the past too, I have always paid her the full price, whether she offered a discount or not. After all, these vendors pay a premium to buy flowers during special occasions. And on that day too, I had intended to pay her the full INR 150 either way.

But there’s the funny thing.

When she first offered a discount and later forget about it, I felt a mild disappointment or something. And it is possible others might have felt the same way in that situation. It is odd, since I was ready to pay full price, right? A loss aversion of sorts.

Trying to quantify my feeling on a Disappointment-Delight scale [-10 to +10] (-10 being very disappointed, 10 being very delighted), I got:

  1. If I go to the stall on a regular day, and I am charged INR 10 for a rose: 0 (on the scale)
  2. When the woman first quoted the festival price of INR 15: -5
  3. When she said she’ll let me buy them at INR 12: 5
  4. When I made a mental note to still pay her INR 15 a piece: 9 (it is always priceless to expect and see smiles on the faces of those who work hard to make a living, when you pay them full price and not be the asshole who squeezes an extra buck out of them)
  5. Her later quoting the original rate of INR 15 a rose: -8
  6. Me then paying her the original rate of INR 15: 3

Despite my intended payment amount and her final quote being the same, my delight level dropped from a 9 (in pt. 4) to probably a 3 (pt. 6).

-End of story-

What do you think? Could experiments/ experiences like this one, experienced by a single person, be then gauged on a list of parameters for other people in the same city, country, and in other parts of the world? The objective not being to find a single global thumb rule or measure (like 70% or 8/10 on delight). But rather, to see how different groups of people fare on each such experiment/ experience. It need not be a labour-intensive effort. Data could be crowd-sourced.

Could this approach be a little less presumptive and a little more accurate than prevailing forms of research studies?

Read about a possible alternate solution that I propose, here: Sample Size of One: Towards a Possible Solution
#SampleSizeOfOne #BehaviouralScience #BehaviouralEconomics
%d bloggers like this: