"Do I believe in Astrology"? This is a question that I used to think about when I was a kid. What could be a possible relation between celestial bodies and human nature when looking from a scientific perspective? How do you really quantify human mood ? It was some 20 years ago, when I read upon a topic in some science magazine that my Mom brought me. It was on the activities of Pineal glands and occurance of flight accidents. Pineal glands are some glands in our body which are called 'biological clocks', since they control our temporal functions like aging, fertilization etc. These glands are set when we are born and they go through some 'activities' with some frequency. The claim in the magazine was that there is some relation between the birthday of the flight captian and the accident dates. Well, that was an interesting thought and seemed to give a possible scientific explanation (atleast for a kid), stating that you are trying to model the state of your Pineal gland (which basically controls your mood) as a function of the position of celestial bodies. I am not very sure how much is this true though!
Well, now I have grown up and have come a long way travelling through time for the past 20 years, and I had more or less rejected any other possible explanations of my astrological relations. Actually, I had stopped believing in astrology, since finding a relation between some biological activities and position of celestial objects was a bit too overwroughted and fancy from an engineering or mathematical perspective. It was a few years back, when this anecdotal event happened. I was in Hyderabad and a friend visited me. It was early in the morning and he was reading the newspaper, when I fancily asked him what was intersting about my horoscope that day! He perused through the paper and started laughing saying that I was going to have a big trouble because of a friend that day! We laughed it across and went ahead with our day! Everything was normal, when towards evening my friend started getting blisters on his face!! He was actually taking care of another friend who had chicken pox a few days earlier! Well, such prognostications happen many a times in our lives and we just passes it aside believing mere coincidence. But is there any deeper meaning to it ? Can we build a theory between astrology and the rest of the scientific knowledge ?
/* advice-- this para is hardcore technical :) you can probably skip to the next section if you don't understand them*/
Now, let me get to my point. My idea is not to see why astrology works, but trying to figure out what could be a possible mathematical framework for it to work. Suppose we have a huge database of the birthday of people and their mood for everyday or say possibly every unit of time. Suppose we have a huge astronomical database of the positions of various celestial objects. Suppose we do a cross-correlation between these two datasets, do a dimensionality reduction and project the data onto the principal axis? More mathematically, we can confine our mood of people to whether they are happy or not; where, say happy is greater than being sad. Later doing some kind of an isotonic regression to map it as a function. Here, the principal axis of projection will provide us with the most important celestial object whose path influences our horoscopes the most. The approach is just what we use in Machine learning for prediction or regression, except that now we are trying to draw a parallel between a scientific technique and some fringe-science. Just like the case that the predictions by machine learning techniques will not give us an absolute answer, we know that horoscopes also will not (Possibly because we are not considering all the directions of variations). People have been collecting astronomical data and associating it with human nature since ages and when the dataset approaches infinity, the relationships become more and more apparant (assuming there is a relationship), and we are seeing this, anyways (looking back at my story) :) .
Ok! So what next? Well, in this world of mass communication and internet, it might not be a big problem to collect the moods of people (let say we just use the status of Gtalk or Facebook :) especially the similies :P ) and getting the birthdays of people is even more trivial. We have an 'infinite' number of people using these programs and a company like Google can very well get this information in no time. Astronomical data is always there since ages. Now, from the above process if some useful pattern comes out then probably we can make better predictions about the mood of people! Hmm... that might be a bit scary! but atleast its nice to think it will work :) Well, it should work, since the horoscopes I belive to some extend does work and "seemingly there is some relation"!
In conclusion, the target of this blog is not to prove anything, but to hint that using Machine learning techniques could possibly put some light on saying that there is something to be believed in horoscopes, and trying to find the exact connection between human nature and celestial bodies is worth a try!!
7 comments:
I was wondering how computer based horoscopes work ? There are a lot of kiosks , esp. at melas and fairs,where they give you your horoscope once you give them elementary details like bday, time of birth ..etc. wonder how that works.
interesting post! actually a more interesting scientific thought behind astrology is the existence of two world views to it... one the western spherical system which is purely based on metrics and distance/direction.. the indian system that is purely based on topology and adjaceny.... of planets as present on a particular day... as the interaction between planets governs the mood of a particular person...
1. Machine learning always finds patterns when you "ask" it to find patterns, even if data is inherently random. AFAIK, no machine learning (or compression) algorithms exist that can tell you when things are random.
2. Humans are provenly incapable of detecting (or generating) randomness. This is a feature not a defect. Humans always look for patterns and come up with theories to describe these patterns. This is the simplest explanation as to why astrology and superstition exists in the first place.
Since humans cant generate randomness, there are computer algorithms that exist, that can beat humans consistently in a chance based game such as rock papers scissors. Interestingly they cant beat monkeys, because monkeys are stupid and can generate randomness. Incidentally it is also the reason why monkeys are not superstitious.
3. so if you do run datamining on this huge database of whatever, u *will* find patterns. This doesnt prove that astrology is correct. You cannot consistently prove that astrology will be right all the time.
Monk
Read: http://web.mit.edu/cocosci/Papers/random.pdf
Nice post John! I certainly agree that humans can't really digest randomness and they try to find patterns in data even if the data were random. But this very fact can bring structure to the way humans think, right?
What if there is a lot of structure in the world and in how humans think? Like I said in the first part of the blog, what if there is some biological activity happening in us that decides our mood of a particular day at a fundamental level? And what if the positions of stars and planets in the sky help you in interpreting it?
For example, you wake up every morning according to some time on your clock. Through so many days, you know that you wake up mostly when the small needle of the clock is between 8 and 9. Here, time is just a "hidden variable", but you try to define your sleeping or waking up in terms of time and interpret it in terms of a physical object, like a clock. Similarly, if you see astrology as a big clock, its not hard to see that your mood might be a function of these astrological objects, which means that astrological objects are not influencing me, but I am measuring my state in terms of these objects. I strongly believe that it is the former part that brings in the superstition component to Astrology. People start believing that the planets are influencing their mood, where as its your body that decides it, but since you don't have a way of finding this out, you try to interpret it in terms of some other objects.
Thus, astrology does not provide us with a concrete answer to the mood of a person at particular time, but it might provide you with a probabilistic interpretation of your mood on a day or things that might happen to you (which might be interrelated to something that is more fundamental to you or to somebody else).
Thus, my motivation for the blog was to see astrology in the light of things I do and understand!! And if we can find such a relationship, then probably it might be a good idea to believe in astrology, since they measure something which is 'valid'. Now, what is this 'valid' thing might be a further study!
Very interesting blog. I have also thought about this with interest. For the record, I am closest to being an agnostic (with whatever logical bias that induces).
The thing that most interests me is preference bias that exists amongst people. Believers have no proven reason to be believers - I agree. Similarly, atheists have no proven reason to be atheists. However, an atheist argument is virtually always biased towards showing how your experiment would fail - even before an experiment is conducted. Even believers' arguments are biased, but atheists always seem to get back at believers with this point.
Trying to predict stuff would be interesting. It does not prove astrology to be true, I agree, but who cares. We are talking about statistical predictions ! If planets do seem to predict moods, hell, lets make money out of it. If not, its a failed expedition, like another class homework !
On a more serious note, not finding patterns also does not prove that astrology is false. For a certain occurrence to happen, we need to sample it well enough. We could just be missing patterns by chance, given the infrequency of our sampling.
Essentially, I belive that both believers and atheists have biases, so no one's arguments should be taken seriously. And now you can decide how my agnostic bias has also influenced my comments :)
Neat stuff. I was looking to see if anyone had done this experiment yet when I found your interesting post.
Although machine learning will try to find patterns even if the data is random, you can also determine how strong the prediction capability is with a lot of precision.
In other words, if machine learning could use a huge data set to predict something about human's daily experiences, it could also tell you how certain it was.
I think a really interesting hybrid application of this would be to use big data (every internet user has A LOT of data about their behaviour out there) to do forecasts of our near future. Something akin to weather forecast.
Even if it could only predict a general temperament a few days into the future, that would be EXTREMELY valuable.
This is something investors would love ;)
Post a Comment