I have been a member of Netflix for a long time now. Getting decent recommendations becomes important after one has watched 500 movies. So, in the hope of getting good recommendations, I faithfully try and rate movies that I have watched. The rating system requires between 1 and 5 stars. If you have followed the buzz around the Netflix prize, you will know that Netflix is very confident about the statistical underpinnings of their rating software and I have no doubt they have done a good job.

But, here is my problem, what if the ratings data is flawed. My ratings for a movie depend on a lot of factors, the context I watch that movie in. My ratings are only valid for that particular context. If the person sitting next to you keeps yawning through a movie, you are much less likely to like it than if you had watched it alone. Laughter is infectious too, bad comedies are likely to get a great rating, if the person next to you laughs out loud for every bad joke. Watching Monty Python with someone who does not quite get it can be a drag. Moods play a big part too, I am much more likely to enjoy a scene where a bad guy gets his teeth kicked in when I am a little peeved myself. The same scene would be abhorrent if I just finished reading the latest new age essay on compassion.

My ratings are very inconsistent. Capturing the context is hard, but it seems like it is an important piece of information in the ratings game. This is true of almost all ratings. Tirerack does a nice job of having reviewers add information about their driving style, but more is needed.

Joe has an interesting link on procrastination . Very appropriate for someone who is starting a blog after thinking about it for a long time. The author even has a formula for procrastination, but I wonder how one quantifies the variables – “the desirability of the task” is one such. A question like – on a scale of 1-10 how much do you want a blog, is much more difficult to answer than a question like – on a scale of 1-10 how much do you want a Porsche. The answer to the second question is the same at any point in time, the answer to the first one depends so much on my context on that day/hour.