Our own Marko Rodriguez of LANL and Knowledge Reef is heading in an interesting new direction, to use the power of computers to make us happy. Below is an interview with him. It raises several issues: 1) He seems to be proposing a data mining approach to determine what makes people happy and a system to improve decision making and suggestions so more people make good happiness enhancing decisions. Having worked for many years with data mining systems, I can see a number of practical and theoretical problems. For example, how to define and measure the many dimensions of happiness, how to get the actual data, how to identify and evaluate all the options for a decision, how to balance short term and long term happiness, and so forth 2) He proposes that use is a good way to determine if the system is actually effective. That is, if people use it then it meets its goals of increasing happiness. That does not seem to me to be an intellectually satisfying way of determining if a system is effective in meeting its purported goals. Is there a better way of determining this? 3) He makes an important distinction between frequent decisions such as which restaurant to eat in, and once in a lifetime decisions such as who to marry (for most people) or where to go to college. He recognizes the data problems of evaluating the once in a lifetime decisions, but seems confident that simply looking at many such decisions will give his system all the data it needs. Im a skeptic. Having looked at billions of decisions for frequent actions, e.g., soft drink purchases, you quickly find out your decision matrixes are actually pretty sparse when you try to account for many of the key variables, and your statistical models based on these do not always have the accuracy you need. Expanding this to once in a lifetime decisions adds orders of magnitude to the difficulty of the problem. Jack Stafurik Computational <http://ieet.org/index.php/IEET/more/fe20090816/> Eudaemonics: Expert Happiness Systems Marcelo <http://ieet.org/index.php/IEET/bio/rinesi/> Rinesi Frontier <http://www.frontiereconomy.com/2009/08/marko-a-rodriguez/> Economy Posted: Aug 16, 2009 This is an interview with Marko A. Rodriguez, a scientist at the Los Alamos National Laboratory. Besides doing basic research on applied mathematics and computer science, he is doing work on computational eudaemonics the use of computer algorithms to increase happiness by helping us make better decisions, even suggesting new options. Do you think the widespread use of eudaemonic algorithms will be contingent on the embracing of an Aristotelian ethic and concept of happiness, or is their usage compatible, in the sense of there being a potentially strong demand for them, with the contemporary ethos? I think the concepts of Aristotle, Norton, Hobbes, Flanagan, and even Rand (to some extent) are all barking up the same tree. And while that species of tree may be the same, the individual instances of it will be different. That is, each person will have to find their own eudaemonic path, where the role of computational eudaemonics is to support the individual in this discovery process. Moreover, for these algorithms, its a process too. Some will live and some will die in this society of algorithms, but the society will continue to evolve and adapt to the human condition. When individual algorithms work well with the human and the human allows them to work, then computational eudaemonics will be serving its purpose. How do you see the process of research and validation of an eudaemonic algorithm for decisions with impact over the long term? With respect to validation, I believe the answer to that is the answer to this: do people use it? Take Google for example. There is no formal proof that PageRank is a good algorithm to rank webpages. However there is a pragmatic proof. The pragmatic proof is the fact that people use Google regularly. Similarly for a eudaemonic algorithm, if it survives to be used another day, then it is good it is valid. Do you feel we have or will have enough information in this generation to data mine patterns about, say, the number of children a couple might want to have? I think what will happen is that more and more data will be exposed in the Web of Data. At first, it might just be a better recommendation algorithm but with enough information in the Web of Data and enough insight on the part of the algorithm designers, we may just end up putting more faith in the algorithm. p>Theres a clear profit motive for a search engine or a retailer to create a good algorithm we interact with them often enough to infer their quality and either become repeating customers or shift to a competitor with a better algorithm but for decisions take very seldom (e.g., choosing a major), there would be both a great demand for good algorithms and an unclear process by which the worst ones could be filtered out. What are your thoughts about who might come up with eudaemonic algorithms and their motivations?p>As you note, there are recommendations that are based on repetition: e.g., movies, books, music, webpages, etc. And, as you say as well, there are one time only recommendations: e.g., which major to choose in college. However, you can see these one time only recommendations as happening in repetition not through the individual, but through the population. While the one time algorithm may be faulty for an individual at a particular point in time, it may be gathering enough data points to be successful for the next individual down the line. I think the saying is: Rome wasnt build in a day. I dont know how accurate these algorithms can get, but there is a sense of better and worse. Moreover, we understand, to some degree, why we like certain movies, books, ideas, etc. So, being able to represent those biases computationally may bring us beyond recommendation and into a world of eudaemonia. For further information: Rodriguez, M.A., Watkins, J., Faith in the Algorithm, Part 2: Computational Eudaemonics, Proceedings of the International Conference on Knowledge-Based and Intelligent Information & Engineering Systems, Invited Session: Innovations in Intelligent Systems, eds. Velásquez, J.D., Howlett, R.J., and Jain, L.C., Lecture Notes in Artificial Intelligence, Springer-Verlag, LA-UR-09-02095, Santiago, Chile, April 2009. [http://arxiv.org/abs/0904.0027] _____ <http://ieet.org/index.php/IEET/bio/rinesi/> Marcelo Rinesi is the Assistant Director of the IEET. Mr. Rinesi is Editor-in-Chief of Frontier Economy <http://www.frontiereconomy.com> .
============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
