Sunday, August 26, 2007

Computerized Forecasting and Discrimination

In my MBA class, I learned a quantitative technique called Discriminant Analysis. Discriminant Analysis is useful for making prediction, e.g. whether a person will buy the latest Harry Potter book or consume organic food.

In my assignment, I devised a model which predicted if a student would be a high scorer in examination. Prediction was made based on a number of questions asked, such as:

  • How much time a student spends on doing homework?
  • How much time he/she spends on sports?
  • How much time he/she spends on surfing the Web?
  • How many siblings does he/she have?

My classmate devised a model to predict if a person would be a potential terrorist. Here is a concern: If a person is predicted to be a terrorist, will he be subjected to unnecessary spying? Will computerized forecasting lead to discrimination?

In her article Forecasting human behaviour carries big risks, Christine Evans-Pughe writes:

The Surveillance Society report from the Information Commissioner’s Office outlined worries about predictive social sorting on the grounds that could amount to discrimination, create new underclasses and that by totting up of negative indicators from health, school and other records, a predictive model could make its own worst predictions come true. “For instance, if your parents both have criminal records or you have a bad school attendance record because of poor health, even if you are the best-behaved kid in class, you will find that every teacher is likely to treat you with suspicion."

As a conclusion, Evans-Pughe sums up:

Computerized forecasting techniques are certainly useful for stores, but flawed when it comes to complex human issues.


  1. Oh doing your MBA.... Are you working part-time and studying part-time?

    I guess models & templates for studies can only work as a reference. It can be a guideline for a lot of things. But sometimes it is full of crap. Even business models & forecasting models in my company gives me the headache and I end up using my own....

    Hehe, I am a PURE science student working in a business consulting company, I totally got lost seeing all the business abbreviations such as CAGR, GDP... I dun even know what the heck it means... Hehe.. perhaps I should do MBA :)

  2. Princess,
    I work full time and study part-time.

    I am thinking: if I live in the U.S., the 'profilers' probably would predict me to be a mass killer. Hahaha...

  3. Muahahahahaa.... i know what you mean about the 'profilers'. All these tests sometimes the outcome of the result (by statistics) are so damn funny. I do feel like I am a psycho some of the times. So you are in Malaysia at the moment?

    Work full time and study part time ar? No more weekends for you lor... And let me guess.. No gf either? :P

  4. princess,
    Yes, I am in Malaysia. My classes are mostly in the evening, so I still have a little bit time in weekends.

    As for your last question... you can find out in my blog.

  5. Those profilers are pretty interesting though I do agree with Evans-Pughe's summation on the techniques being flawed when it comes to complex human issues.

    Also, I think these forecast results may do more harm than good to people who are not strong enough and could succumb to the ill-effects of auto-suggestions. Thank goodness, not many people take these results seriously or am I mistaken?

  6. Hi happy,
    Your point of auto-suggestions is worth pondering. So if a kid is predicted to be a not-so-bright student, he/she probably would simply give up studying. That would be bad.

    Mmm... the banks are probably serious about the forecast of potential defaulters.

  7. forecast? believe them when they sound so good to be true only! human nature...

  8. kai,
    A forecast sounds good if it is simple. But a simple analysis could be seriously flawed.

    Yeah, human nature...