Tuesday, 11 March 2014

On Bad Science and False Premises

Many scientists do bad science by hiding behind empiricism. 

Empiricism is when you do things using the scientific method i.e you observe something, investigate it by building a hypothesis based on what you know, look at the existing literature, create a hypothesis, conduct a study, experiment, collect evidence and build a theory. 

Here's the thing. Your hypothesis has to be based on evidence. You cannot simply make stuff up to explain what you see, and then justify your assumptions by claiming that you're testing your hypothesis. You can't do that. 

If you create a hypothesis based on baseless assumptions, and then test it and find data that supports your hypothesis, it might only be correlational, or coincidental, or due to confounding variables. The data itself would not indicate cause and effect, or the truth of your hypothesis. This is because your original theory was made up to begin with. A false premise. 

The data collected in support of a baseless hypothesis might be good, but since it has been collected based on a preconceived notion of its role in the hypothesis, it is meaningless.

"I think men like playing first person shooter video games because of their hunter gatherer tendencies. Let me conduct an experiment to see if I can indicate this to be true. Yup, there's an association between the two variables. Men play more first person shooter video games than other types of games, and they play these more than women do. These games simulate hunting, and in the past men hunted more than women. Therefore, my hypothesis is supported."

Of course, the only reason this supports your hypothesis is because you made up the conclusion to begin with. The data might be sound, and the association might be real (and due to any number of other factors), but the inference is false, because you're merely fitting the data into a preconceived role or conclusion in your mind, and in your silly hypothesis. You've decided your conclusion in advance, and then you're letting your results justify your conclusion, even though they might indicate any number of other things that haven't been hypothesised yet. 

"If A exists, then so must B. I found A, therefore B exists." 

Unfortunately, the former is an assumption. Therefore, the implication in the latter is not always true. Bad researchers don't get this.

You cannot use the fact that you're testing your hypothesis to excuse the fact that your hypothesis is based on no originating data, and merely something you dreamed up. Your starting point must always be evidence. Without evidence, you have no starting point, and no hypothesis. You cannot indulge in guesswork as a replacement for evidence. You cannot assume that a complex facet of human behaviour exhibited by people today has an evolutionary basis originating in a behavioural trend exhibited by people 500K years ago.

This is where evolutionary psychology gets it wrong. Researchers make a guess, a 'just so' story that feeds into our existing status quo of how we imagine the world should work (it sounds true therefore it is probably true). When confronted with the fact that they have no evidence to support their theory, they defer to helplessness and empiricism - 

"No one can prove evolution. No one was around to record behaviour 500K years ago, so my hypothesis could be wrong, but at least I can test it. So my approach is scientific. At least I'm better than a religious person. At least I'm willing to test my theory by creating and testing hypotheses and admit I could be wrong." 

What. A. Crock.

It is extremely important to be consciously aware of your underlying assumptions and implicit biases when formulating a theory. Do researchers do this? Nope. They allow their biases and assumptions to dictate their research questions, their hypotheses. Whether their hypotheses are true or not is irrelevant. Once a hypothesis is created and tested, it becomes part of the scientific narrative. Your students, peers and other researchers will spend the rest of their lives creating hypotheses that match your theory, hoping to reject null hypotheses in favour of ones that support your theory, all founded on nothing, everyone wasting their time and energy.

If someone thought their desk was a dragon, we'd call them crazy. However, if they spent their lives trying to test the animatedness of their desk, believing it were alive, would you still take them seriously? Of course not. But people around the world everyday do the same with bad hypotheses. Hypotheses with no basis. Hypotheses created based on someone's assumptions. Hypotheses created because they 'feel right'. Hypotheses created because their creator lets them reflect his or her own biases about how the world functions.

Your ability to create and falsify hypotheses does not justify your creation of theories with no underlying bases i.e making bad assumptions. It is important not to confuse science with empiricism. You might be a brilliant empiricist, but you'd still be a bad scientist if the assumptions inherent in your empiricism had no basis. It's easy to hide poor assumptions and reasoning behind good empiricism. Don't do this.


Saturday, 8 March 2014

On Music and Memories

What if I scanned the brains of 200 people for activity patters while they listened to various types of music? What if I also got them to create lists of their music preferences, to find correlations between music preferences and brain activity. What if I then worked backwards to connect these patterns to Affect states resulting from other experiences?

Here's what I'm getting at - What if musical preferences are merely side products of neural plasticity and neural responses to other auditory stimuli? You're exposed to certain auditory stimuli as an infant. These stimuli are positively or negatively reinforced through cultural context. What if our music preferences are not merely learned, but conditioned? We are machines after all. I'm talking neural plasticity. The basis of learning. Neural networks strengthen synaptic connections or form new ones over time. They're never static.

And here's where we face the problem with connecting models of learning with models of neuroscience. Nerve cells fire at the rate of milliseconds. Overt learning behaviour models deal with seconds, minutes, hours. But I argue that there must be reinforcement involved in music preferences. You not merely like a piece of music, you learn to like it. If you like Metallica or even beer, we know that it's because your nerve cells that processed that type of sensory input have strengthened their connections over time. My arguement is that it could be because these same nerve cells were responsible for processing other stimuli, that were strengthened for some reason. So these neural networks are either growing more connections, or firing quicker. Or both.

Why would they strengthen their connections? Who knows? Maybe because the semantic feedback associated with these other stimuli led to chemical changes in your body that were favuorable to you. In other words, they made you feel good. Or then again, maybe there is a genetic element. Maybe a group of genes, on being environmentally activated, help reinforce the strengthening of certain connections that favour certain stimuli, or inhibit connections that would have made you despise certain other stimuli.


Friday, 7 March 2014

On Education and Expectations

If you want a career with a top organisation, a good start would be to go to a top university.

The UN, IMF, World Bank, RBI, economic advisor to the PM, the Fortune500, etc. aren't known for hiring University of Mumbai grads. The point I'm making is, the opportunities that you are presented with are a direct function of where you study. You need your own academic lineage to enter the 'big league' employers.

For example, someone with degrees like IIT + IIM + PhD at Harvard, etc. are more likely to be offered a prestigious high paying job than someone with not so hallowed credentials. And it's not necessarily because the better universities offer better academics. It's just because they have better reputations because they have better selection criteria because of a continuously evolving relationship with selected employers.

Now, most people shouldn't be bothered about these criteria. But it turns out most people in india have a 'family lineage' to protect. Which is why you see a correlation between what your dads do for a living and where you study. The more important your dad is, the more he expects of you. Privilege begets privilege, and status begets status. Most guys occupying the top roles in India today come from privileged backgrounds, with pedigreed degrees. Their dads were IAS/IFS, law, politics guys from some of the best universities around, and they want the same for their kids. And so their kids have to go to top universities too. Sure these kids are smart and work hard, but they also have expectations to live up to.

Money plays a role too. And so does culture. Rich people, people with their own large businesses, or people climbing the corporate ladder, want the best for their kids, because it ensures that their kids have access to the best opportunities in the world. Can you imagine some Tam Brahm guy in India studying psychology if his dad works in senior management in a top company?

It's important to note that I'm not referring to getting to go to a better university because you're rich. I'm not focussing on being unable to access the same options as richer folk. That is a serious problem, especially in other countries. But in India, barring extreme economic hardship, I'd say the educational opportunities favour equality. IIT + IIM fees aren't that high.

There are exceptions though. You might be middle class, and still decide to go to a lower ranked university because you need quick employment, which is where how affluent you are comes into play. Perhaps an element of regret sets in, when you realise where you got your degree matters a lot.

However, what I am focussing on is more of a selection bias in where you study, brought about by family pressure and social standing. For the Indian middle class, it's enough that you're studying engineering. For families higher on the social ladder, it must be IIT. And if you can't make it into IIT, that's where the huge bank balance works in your favour - Harvard/Wharton come into play. A relative once told me it's pointless getting a MBA unless it's on the 'ET MBA top 100' list. I've realised it's the same for other Bachelor and PhD programs too.

The better the university, the better your prospects. It's better to get a psychology degree from Cambridge University than Glasgow Caledonian, if you want to do world class research someday. But most people don't want to do world class research & hobnob with people like Dawkins or Baron Cohen, so they don't mind going to Glasgow Caledonian. I'd say where you study depends on your own motivation and what you want, which depends on what your family is like and how successful they are, and the pressure you experience from the expectations they have of you.

The best we can do at this point in our lives is accumulate enough resources to make it easier for our own kids to go to good universities, without ruining their lives in the process, and making good education more widely available to as many people as possible, and trying to weaken the existing system of a strong selection bias favouring recruitment from a few select universities.


Thursday, 6 March 2014

On Crime and Punishment

You don't fight crime by increasing the magnitude of deterrents i.e the magnitude of punishment. Evidence shows that this just doesn't work. Changing the punishment from 5 years jail time to 20 years jail time, or from life imprisonment to capital punishment just doesn't work. You can however reduce crime by increasing the magnitude of what people value in life (which translates to an increase in what perpetrators can expect to lose if they commit a crime).

Deterrents only work if they're mapped to what people value.
 People with more to lose commit fewer crimes. This reflects the changing context of the cost-benefits-risks equation. Getting caught & going to jail means something different to a slum dwelling high school drop out than it does to a middle class Indian, because a middle class Indian has more to lose by going to jail (his freedom, status, job, money, friends, etc.) than a slum dwelling high school drop out (who will end up losing less since he has less of these to begin with and so places a lower value on their loss and is therefore more risk seeking). Governments can use this principle to decrease sexual predation by increasing the risk factors for offenders, thus changing their decision thresholds w.r.t offences.

And how exactly do you do this? By increasing economic development and inculcating moral values in a population. Because most things people value are either material or perceived mental states. And an increase in these things will mean a lower crime rate, since people now have more to lose. And so people who value society's perception of themselves and have more material wealth, status, empathy, etc. will be more risk-averse when it comes to committing a crime, since they now have more to lose, in terms of both material comfort and the way they feel society views and judges them.


On the Nature of Genius

I've been reading up on genius lately. What makes a genius. Who can be considered a genius.

There seems to be little distinction between literary, scientific, and musical genius. It doesn't seem to be what you change, but how you change it. It's more about what you create, and how much of a difference this makes to the world. Genius involves thinking outside the box. Most scientists, artists and creators don't do that. They simply work within a slightly enlarged box, enlarging it from the inside, standing on the shoulders of others. 

A true genius finds a way to look anew at what everyone else sees, and see things no one else does. A new way of thought, a new pattern, a new philosophy, a new logic to how the world works. A game changer. A new type of art, a new type of mathematics, a new way to connect the dots. Creativity and originality. Education helps. you need to know how the existing status quo works before you reject it and establish a new one.

It seems to me that what truly separates a genius from the rest are not their abilities themselves, but associating their abilities with a leap of progress. Newton was smart. So were/are the last 50 Nobel prize winners. That doesn't make all of them geniuses. They all played a part in progress, but only one of them revolutionised the field of mathematics.