Questionable experiment giving AB testing a bad name?

Note that the debate here is not about the validity of AB experiments, but rather about whether it is ethical to do experiments that involve tweaking “psychological” levers.

While the answer may seem obvious at first sight (don’t play around with peoples emotions!), let us be clear that the issue is far more complicated.

Have you stopped and wondered how every ad you see online or on TV or on a bill board tries to manipulates your emotions? Have you considered how news anchors and news channels put their own spin while reporting any news? Have you noticed how every organization with an agenda will manipulate a news event to suit its needs? Even in our day to day life, our kids, our friends, our spouses all can and do manipulate our feelings. This is the way humans are! We are highly emotional and suggestive creatures and using emotional manipulation is part of our survival repertoire.

So before jumping all over Facebook for doing such an experiment, lets step back and make sure our response is a measured and calibrated one.

By the way, I do not own Facebook directly, and I am not even that heavy a user!

causality, Linear Analysis, predictability

When a Health Plan Knows How You Shop – NYTimes.com

June 29, 2014 anandoka Leave a comment

When a Health Plan Knows How You Shop – NYTimes.com.

Predictive analytics at work! Apparently if you are a mail order shopper you are more likely to use emergency services.

Be that as it may, we should guard against the cardinal sin of confusing correlation with causation. As the article itself points out, you may be home bound because of an infirmity or an inability to drive due to a medical condition and that may also be the reason you are more likely to need emergency care. So in such a case, mail order shopping would only be a symptom, not the cause, and stopping mail order shopping may not make you more healthy. On the other hand, doing a lot of internet shopping probably also involves a lot of sedentary activity, and that indeed can cause health problems. So perhaps switching to mobile internet shopping while on the move may actually be a helpful change.

Disentangling causation from correlation can be a nightmare. And it needs more that statistics and machine learning. It needs some aspects of control theory, time series modelling and good old physics. ML just tries to build the model that gives the best precision/recall trade-off over a set of validation data. But for understanding causation, we need to use models that are not only precise, but also understandable and plausible in terms of physical processes.

My favorite example is the phenomena of inter symbol interference (ISI) in wireless communications. We can fit all kinds of complicated models to the observed interference corrupted signal including complicated kernel methods and even deep learning. But as it happens, the physics of the situation indicates that ISI is a relatively simple linear phenomena involving convolution of the transmitted signal with a “channel response” that has several distinct “multi-path” components. Armed with this knowledge we can easily solve the problem of identifying the model using a tap-delay-line model and simple least-squares methods. But importantly, the learned model corresponds to actual physical reality and each “tap” in the model actually maps to a real physical reflector of the radio waves. The strength, phase and delay of the tap can tell us something about the physical object that is reflecting the radio waves. (See my previous blog.)

The moral of the story is that while probability and ML is important, we should not lose sight of more classical disciplines like control theory and mechanics, if we want to go beyond prediction, and actually find the cause of interesting phenomena!

data analytics, feature extraction, probability and information

With light echoes, the invisible becomes visible — ScienceDaily

June 19, 2014 anandoka 1 Comment

With light echoes, the invisible becomes visible — ScienceDaily.

This is a really cool invention, and especially speaks to me!

I have worked with with wireless technology a lot, where the electromagnetic transmissions are in the millimeter or cm wave length regime, and there also we see this phenomena called “multi-path” propagation. Signal from the transmitter reaches the receiver via numerous paths due to reflections (often when there is no direct path at all), and each component therefore has a different gain (amplitude and phase) as well as delay (time of flight). (It is quite normal to see hundreds of paths in cell phone signals for example.) Getting all these multiple paths properly “aligned” so that the signal can be easily demodulated, is the principal work of a wireless engineer and traditionally this is done using techniques like self calibrating equalizers and interference cancelers.

Interestingly there is another modern approach where we “leave the signal alone” and instead use a complex decoder that uses probabilistic methods to decode the information from the multi-path signals. In this case the decoder becomes complex and requires a lot of computational power.

Especially in ultra-wideband radio the multipath signal carries very rich information about the physical location (like a fingerprint) and hence can be used for location detection by using complex signature-matching algorithms. (This is something I have worked on quite a bit). In this case it is not the information modulated on the signal by the transmitter that is of interest, rather it is the information modulated on the signal (by multipath phenomena) by the environment that is of interest. I believe this current invention is very close in spirit to this latter approach.

What is very impressive is that they are doing this for visible light frequencies rather than millimeter or microwave frequencies, and hence the “fingerprint” they are developing is in fact a visible image! The fact that it allows one to “see around a corner” is a great application of the technology of course. But it need not be limited to that. I think the technology could also be used for more things like medical imaging, or for detecting micromotions in an area (say to detect mechanical vibrations or earthquakes). I can think of quite a few other use cases too 🙂 …

In any case this is fantastic engineering, and I am happy to see some of the researchers are from my alma mater (University of British Columbia)!