Theorized (rational) choice process being modeled: doesn’t mean we assume that all humans are rational- we just have a rational standard. So using models we can see the degree to which people act rationally.Examine all possible plans we model people as though the examined all of the possible choices they could make, we’re not saying they did that- we’re just modeling them that wayIdentify all possible outcomes we look at all the possible outcomes of a situation and assign all of the outcomes with values.Value each outcome you place a value on each outcomeEstimate probability of each outcome 20% prob you get offered a crappy job, 10% outcome no job, ect….Choose outcome with highest “expected value” (ie, value x probability) the rational choice is to choose the plan of behavior with the highest expected value, which is value multiplied by probability.EV = Σ Pi Vi (for i possible outcomes) the ‘I’s are just counters, they refer to the number of possible outcomes. If you have 5 outcomes you have to do the multiplication 5 times. It’s the sum of the productsExpected value: value x probability the probability is where the word ‘expected’ comes from. If you think that saying hello to you would give me $10 and there’s an 80% you’ll say hello the expected value of the in interaction is $8.Probability =0 to 1 (Decimals or percentages)1) Avoid reasoning error of assuming that all outcomes are equally likelyLottery: two outcomes, so 50% chance of winning people think that that are two things that could happen, I could win or lose. FALSE there’s 1 in 300 chance you get your money back, 1 in 3 million odds you win.2) Gambler’s fallacy: being “due” for a red card, etc. if you flip a coin and it comes up heads 3 times in a row. What the heads its heads on the 5th toss? STILL 50%. So don’t think that because you’ve been losing that you’re going to win eventually- its still the same oddsDecision trees: merely a graphical aid to seeing which outcome has maximum EV these are teaching aids. Go look at the example of expected cost of a car. So the expected cost of fixing the car is $208, or $175 for a car that 100% will work.. the rational choice is to buy the roommates car.Draw out the whole tree, including all the contingent branches once you’ve drawn them all the way out to the right, you do math all the way back to the left and attach values to each thingThen start collapsing to the left, until you end with only the branches representing the first choice the basic choice was buy your roommates’ car or repair your careThe largest EV predicts the first choiceContinue the process:Given the first choice, collapse back to the next decision (ie, second choice)Alternatively, can just multiply probabilities along each branch, and list the product at the end of the final branch point if you don’t like decision trees, you can multiply the values for each branch, ect..If a thing can happen N ways, add up all N probabilities to estimate its likelihoodUtility: an all-purpose metric $15 off may not get a millionaire into a restaurant, but it may convince a college kid. Money doesn’t mean the same thing to everyone. So we need an all-purpose metric. Dollars are inconvenient and misleading- but there can be some translations so everyone is on the same metric.Can ask people to rate value of $100, a puppy, and a quick meal on the same scale, maybe 1 to 10. Teacher offers to hire someone to kill a puppy. Almost everyone has a price that they would agree to do it. So everyone has a place on this scale. We can translate puppy killing to dollars. So the point is that you can take a thing and make a value of it. WE USE A METRIC OF UTILITY because a dollar is meaningful. So we will talk about how many ‘utils. We may need an anchor point. It’s a nice neutral scale.When ask people to make their own probability estimates, call these “subjective probabilities” rolling a dice is an objective probability, when looking a humans, it is subjective probability. Everyone right now can give a probability that they will get a good-paying job when they graduate. If we have an estimate of the persons utility and probability . to predict a person’s behavior we have to figure out their subjective probabilities, ex. “whats the probability you are going to graduate?”Get SEU models – subjective expected utility not in textbook- called expected values in the bookNotice normative character of EV and SEU models:Predict rational choice, given the starting info we DON’T assert that humans are rational. We assert that we have a rational model that predicts what large groups of people will do. We can hold up the rational choice next to the actual choice. People think about what the rational choice and they have built models that are based on a normative standard and logical grounds, we haven’t assumed that everyone is rational- we just have a standard to measure people against.Other social science is descriptive, which is predictive we evaluate the descriptive work, was it a dumb choice or a great choice? We can tell the difference because we have a normative standard. We can do this with models that come out of the choice frameIf people are rational in some domain (testable), then same model can be bothAggregation: or the law of large numbers, sometimes applied in order to use a normative model for prediction in the book they focus on per say, one rat, but in practice we want the ability to predict groups. To predict group behavior its appropriate to say that 80% will do this, 20% will do that. You can make accurate predictions about accumulations of peopleAssumption is that errors (irrationalities) are random and cancel out with a large enough sample law of large numbers, people themselves are squirrely as hell, so individual people are hard to predict. There is a lot of random error in human behaviors, but the law of large numbers says the errors will cancel out. The more data you get, the more you will settle on the correct answers/predictions-Key thing is that one choice has punishing results, instead of one just being better than others. Math the same. Psychology of it may not be. The math of decision with risk is the same, but the psychology of it may not be. Ex. Drug that would raise your IQ by 20 pts 99% of the time. The math works out that its rational for you take the drug, because there’s a 1% chance you will die- and that’s a big decision even though