Welcome to research – part II

Research as Mindset
As mentioned in part 1, the most important element of doing research is your mindset. Observe a young child well before he begins to understand the ways of the world. Look at the way he/she would go and touch anything that attracts his attention (almost everything in sight, actually!). He would ask you to explain what it is. He would try to name it, perhaps, based on names he has heard so far. You would patiently correct him, perhaps an infinite number of times. He, does not hesitate, to ask you stupid questions – questions which looks stupid to you!As the boy in the emperor’s clothes story, he has no concern of important person or otherwise, and can tell whatever comes to mind. In some sense, this is the researcher’s mindset.

Though not very often perhaps, you would have read papers, articles and occassionally books on various topics. When you find some statement which puzzles you, what do you do? Most of the time you may console yourself: I dont understand this area well enough to follow this. Do you consider the possibility that the author could be wrong? Or there is some mistake or ambiguity in the writing? This ‘respect of authority’ often takes us far away from research; we never stop to ask questions, or explore our doubts. The authors must know better – after all they have written a whole book!If you look at carefully, most books contain errors – irrespective of who wrote it. Many research papers have serious flaws in the quality of writing and the quality of work reported – it does not matter if you are reading an IEEE paper or ACM paper or a local conference paper. The latter has, usually, far more problems, that is all.
So when you read any material, always keep a pinch of doubt always. Never believe everything that is said or written. When inferences are drawn, ask the question – does this conclusion really follow from the base given? Is this much data adequate to draw such conclusions? Is there something the authors are missing? Does the result intuitively make sense? If not, where is the problem? Is the approach adopted fair? Is this the most appropriate under the circumstances? Are all the possibilities accounted for in the data set? Is the data set representative of what is likely to arise in the domain? And so on. There are many questions that a piece of writing is expected to answer. Some of the answers you can find from the paper itself. Some can be got based on your knowledge of the domain. Some may require further exploration. And there lies the key.

Asking these questions keep you alert on one side; your absorption of material is much better and more lasting, since there is a conscious processing of the information being given. The questions prevent you from accepting things on its face value, and hence you are more likely to pick things that are important and valid. You can, therefore, justify your belief in a claim or observation – since you have understood, and not memorised. As a researcher it is your responsibility to do this critical review, since that is what helps the author to refine his work – just as you would want others to look at your work, and give you honest feedback.

This mindset also has a related aspect – attention to detail. Research is not a shallow work. Things make sense only every aspect is attended to. A small bug in your program can lead your result completely astray. So careful testing and confidence building on any program is important, to save embarassment for you in the long run. Similarly, careful organisation of your experimental work, properly labelled and structured is essential. Otherwise, in a while, your research work will be like a directory containing a thousand files of different types. You will be lost in it, everytime you need to find something there. What looks very obvious to you today, wont even look familiar six months from now. So, yes, label the files and data right now – tomorrow may be late!
Research results and data has a lot to tell you – provided you know and want to listen. Learn to listen to them. Keeping them properly organised is an essential ingredient to this; otherwise, the talk will be in a language you wont understand.

Consider the two tables below:

[field-1][field-2]                     [sex  ][pregnancy]

[1           ][1            ]                    [m    ][yes                ]

The left table is a silent table; anyone can pass off that data – looks pretty ok. But the table on the right will hit you straight. You realise there is something fishy. Both the tables represent the same information as far as your program is concerned. Your eyes are indeed quite powerful to detect anomalies – use them. Just organisation is not enough – you need to train your eyes to look for problems. A number of tools for data visualisation are available – they can make complex relationships visible to your eyes, through suitable transformations, etc. But, much can be done without them.

Tables and graphs are often used in technical communication – but rarely used effectively and properly. A number of small things make these powerful tools ineffective. Do you always label the x and y axis in a graph? You must always. Also ensure that the units are part of the label (or visible in some other way). Show grid-lines at major intervals – we are not particularly good at scanning lines even vertical and horizontal. Similarly, look at the graph – rather listen to the graph. Are the values at the various points as expected? A graph as in figure below is trying to tell you that there is a problem – but one can find many papers which does not even mention this anomaly, let alone explain it. The anomaly may be accounted against many different reasons. It could be some error in the output measurement. It could be some corrupted data in the input – which you should have detected, but possibly never checked!It could be a nasty bug in your program, and fixing it may change other values as well!It could be an important result – may be there is an interesting behaviour at the anomaly point, in which case, you should study that area in more depth with more experiments. Another thing to look for is the graph value at common sense points – usually x = 0, y = 0, x = infinity, y = infinity. The values at these points are usually predictable, and hence provides a way to cross check your methodology.

You can see that this is not just for research. These problems can be noticed in your own reports and even work. Such observations, generally called sanity check, is a great tool for software developers and researchers, and pretty much everyone. When you check out of a super-market, if you keep an eye on the prices of items you take, you can make a reasonable guess at the total cost, without using a calculator. So, if the guy at the exit takes 1000 instead of 800, you can suspect something odd; rather than waiting to check the total after you reach home.

This example takes us to another interesting tool for researchers – order of magnitude reasoning, and qualitative reasoning. Do you know how long it takes for a litre of milk to boil on the gas stove? You wont know if it is 2.2 min or 2.3 min, but you know it is more than 1 min and less than 10 min. So you can go and open the door, with the milk on the stove, but you wont go to take bath!

When you find the average salary of a pool of people is 10000, when almost all the numbers are in the range of 1000-2000, you suspect something wrong, and perhaps discover, a wrongly entered salary figure of 1 crore!

The lack of trust mentioned above is not for others alone – if at all, they apply more to your own work. You may be a good programmer; but dont trust your code too much. And if you are getting very good results for an experiment, check it very carefully – life is rarely so good. And if your results are quite unexpected, dont blindly discard the results. May be, there is some interesting phenomenon you have overlooked. Remember the discovery of pencillin, and many other stories of scientific discovery.

Thus, key ingredients of a research mindset are curiosity to know and find out, no hesitation to ask and question, and attention to detail. Be a child – you will enjoy. Of course, as Dr Phatak once mentioned, dont run around without clothes….


4 Responses to “Welcome to research – part II”

  1. Sunil Mandhan Says:

    Very nice article, i really liked it and appreciate your community service.

  2. Sankalp Bagaria Says:

    Very useful article…..Looking forward to next posts….

  3. Chetan Says:

    Very nice article. I quite like the bits where you suggested
    * spirit of inquiry
    * Analysis and not just observation ( “listening to your graphs”. )
    Somehow most of our academic life is indeed spent observing and emulating, than analyzing at all. sigh.

    In that context there were two additional things I have experienced myself
    * prototypes _need_ to be destroyed to enable a good design
    * Analysis should have a definite exit ! else I faced analysis paralysis. 🙂

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: