On Sun, Jan 19, 2014 at 08:38:59AM -0800, Edgar L. Owen wrote: > Russell, > > No, rational agents are NOT "entirely predictable". And the definition of a > 'rational agent' is not someone who always makes the "best choice".
See http://en.wikipedia.org/wiki/Rational_agent > > First of all there is no such thing as "a best choice", because best choice > is a judgement that depends on the value scale of some observer, and it Of course. This "value scale" is called the "utility", and is known by both the agent and the predictor. If the agent can calculate the action with optimal utility, then so can the predictor. Of course if the agent's utility was cryptic, then it wouldn't be particularly predictable, but that situation is unusual, to say the least. > also depends on what time frame the effects of the choice are considered > in. A good choice in the sense of immediate favorable consequences can have The utility is given. It may include a discounted factor for future effects, but once given, the agent is bound to choose the optimum action according to that utility. > disastrous unforeseen consequences down the line. There is no such thing as > a "best choice" because causality is an enormously complex interconnected > network rather than just A causes B. There are always a huge network of > causes leading up to every decision, and huge ongoing network of effects. > > And rational agents are not entirely predictable for several reasons. One, > it is impossible for any agent to know everything about everything, and all > knowledge is observer conditioned anyway (it's different for every > observer). Rational agents are entirely driven by the environment, which is known by all players. > > Second, there are random processes occurring internally to every agent > unknowable and unpredictable to any outside observer that are involved in > reaching decisions. Rational agents do not have internal random processes. Third it is impossible for any outside observer to > entirely know the internal valuation process by which an agent internally > calculates its actions. > The internal algorithm doesn't need to be known. The problem is entirely determined by the agent's utility, and its environmental input, which is known. The predictor can use any correct algorithm to compute the rational course of action. > Third, agents make decisions not on the basis of a common agreed upon > external reality, but on the basis of their private internal mental models > of that reality, and these are always private to the particular agent, and > differ widely among agents. > If by private mental model, you mean information based on past environmental states, this information is also known to other agents in that environment. If you mean some other sort of private mental model, such as religious beliefs, then you no longer have a rational agent. > So rational agents are not necessarily predictable, and of course neither > are irrational agents either. > > Your statements are based on an extremely simplistic view of reality is > which everything is deterministic, and everything is known, and there is > some absolute notion of best choice. Not one of these 3 assumptions is > actually true... > I never claimed that rational agents actually exist in the real world. They are at considerable disadvantage compared to irrational agents in the messy, real-time, and stochastic world we live in. They are, however, and important theoretical concept in economics and agent based modelling. Perhaps overly important. > Edgar > > > > On Sunday, January 19, 2014 12:39:27 AM UTC-5, Russell Standish wrote: > > > > On Fri, Jan 17, 2014 at 10:40:51PM -0600, Jason Resch wrote: > > > On Fri, Jan 17, 2014 at 6:33 PM, Russell Standish > > > <[email protected]<javascript:>>wrote: > > > > > > > > > On Fri, Jan 17, 2014 at 04:08:08PM -0800, Edgar L. Owen wrote: > > > > > Russell, > > > > > > > > > > PS: On second thought maybe we don't agree completely. Though free > > will > > > > is > > > > > quantum random based (we agree on that), it doesn't mean that it is > > > > > "irrational". > > > > > > > > > > If human actions and the actions of other biological organisms > > weren't > > > > > basically rational they couldn't function or survive in the real > > actual > > > > > world they live in.... > > > > > > > > That is not true. Read up on the concept of Machiavellian > > > > Intelligence. A modicum of irrationality is just what it takes to get > > > > ahead in the world, it makes one less predictable to one's > > competitors. > > > > > > > > > > But to be "effectively unpredictible", one doesn't need a truly random > > > source, rather only a small number of bits that remain undisclosed to > > > outsiders. > > > > > > > That is true, but the irrational/rational distinction doesn't lie in > > the same place as deterministic/indeterministic. > > > > Rational agents are entirely deterministic and predictable, but it is > > certainly possible to get deterministic irrational agents, and even as > > you argue, deterministic unpredictable agents. > > > > -- > > > > ---------------------------------------------------------------------------- > > > > > > Prof Russell Standish Phone 0425 253119 (mobile) > > Principal, High Performance Coders > > Visiting Professor of Mathematics [email protected]<javascript:> > > University of New South Wales http://www.hpcoders.com.au > > ---------------------------------------------------------------------------- > > > > > > > > -- > You received this message because you are subscribed to the Google Groups > "Everything List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/everything-list. > For more options, visit https://groups.google.com/groups/opt_out. -- ---------------------------------------------------------------------------- Prof Russell Standish Phone 0425 253119 (mobile) Principal, High Performance Coders Visiting Professor of Mathematics [email protected] University of New South Wales http://www.hpcoders.com.au ---------------------------------------------------------------------------- -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/groups/opt_out.
