He speaks here. He raises a number of issues with self-reported happiness. An example he gives is someone moving to California because he thinks he will be happier there. If you ask him if he is happier in California, he will immediately think of the better weather (compared to, say, Ohio) and say he is happier. However, in reality there could be all sorts of other things that make him worse off in California, and many measures other than self-reported happiness would indicate that he is unhappy.
At the end, prodded by the moderator, Kahneman says that policy makers should pay attention to happiness research. I think that we are not even close to being at that stage. Frankly, Kahneman strikes me as only just starting to grapple with some really difficult philosophical and methodological issues with happiness research. How can he start the talk by saying that the term "happiness" is ill-defined in the research and then end up by saying that policy makers need to pay attention?
The study of happiness strikes me as being in 2010 what Freudian psychology was in 1950 or Skinnerian psychology was in 1970 (I may have my dates somewhat wrong, but I am trying to remember when they were near their peak of acceptance by intellectuals in general.) In retrospect, those earlier fads were based on shaky philosophical foundations and an empirical strategy that was focused entirely on confirmation. We look back on Freudians and Skinnerians and find it hard to believe that they could have been so blind.
Perhaps happiness research will evolve in a more robust manner, so that the analogy with earlier fads will prove false. Until that happens, I would worry that if you implement a policy on the basis of happiness research today, then there is a very high probability that subsequent analysis will show that the policy was mistaken. Some day, we will look back on happiness researchers and find it hard to believe that they could have been so blind.