Traditional Approach . . . A Testable . . . The Above Testable . . . For Close Alternatives, . . . Asymmetric (Libertarian) Maybe Human . . . Paternalism: Explanation How to Take Into . . . Another Case when . . . Based on Decisions Under Asymmetric . . . How Does Our . . . Interval Uncertainty, and Potential Applications . . . Possible Applications to Home Page Title Page Education ◭◭ ◮◮ Olga Kosheleva 1 and Fran¸ cois Modave 2 ◭ ◮ 1 Department of Teacher Education 2 Department of Computer Science Page 1 of 12 University of Texas at El Paso Go Back El Paso, TX 79968 olgak@utep.edu, fmodave@utep.edu Full Screen Close
Traditional Approach . . . A Testable . . . 1. Outline The Above Testable . . . • In general, human being are rational decision makers. For Close Alternatives, . . . Maybe Human . . . • However, in many situations, they exhibit unexplained How to Take Into . . . “inertia”, reluctance to switch to a better decision. Another Case when . . . • We show that this seemingly irrational behavior can be Asymmetric . . . explained if we take uncertainty into account. How Does Our . . . • We also explain how this phenomenon can be utilized Potential Applications . . . in education. Home Page Title Page ◭◭ ◮◮ ◭ ◮ Page 2 of 12 Go Back Full Screen Close
Traditional Approach . . . A Testable . . . 2. Traditional Approach to Human Decision Mak- The Above Testable . . . ing: A Brief Reminder For Close Alternatives, . . . • Situation: we have alternatives A 1 , . . . , A n . Maybe Human . . . How to Take Into . . . • Idea: alternatives are characterized by their “utility Another Case when . . . values” u ( A 1 ), . . . , u ( A n ). Asymmetric . . . • Preference: A i is preferable to A j if and only if How Does Our . . . u ( A i ) > u ( A j ) . Potential Applications . . . Home Page • Empirical testing: we need to compare Title Page – empirically “testable” behavior (such as preferring ◭◭ ◮◮ one alternative A i to another alternative A j ) and ◭ ◮ – difficult-to-test comparison between the (usually un- Page 3 of 12 known) utility values. Go Back • Conclusion: empirical testing is difficult. Full Screen Close
Traditional Approach . . . A Testable . . . 3. A Testable Consequence of the Traditional Ap- The Above Testable . . . proach to Decision Making For Close Alternatives, . . . • Fact: for every two alternatives A i and A j : Maybe Human . . . How to Take Into . . . – either u ( A i ) > u ( A j ), i.e., the alternative A i is bet- Another Case when . . . ter, Asymmetric . . . – or u ( A j ) > u ( A i ), i.e., the alternative A j is better. How Does Our . . . • Comment: exact equality of u ( A i ) and u ( A j ) is highly Potential Applications . . . improbable. Home Page • In the first case u ( A i ) > u ( A j ), Title Page – if we originally only had A i , and then we add A j , ◭◭ ◮◮ then we stick with A i ; ◭ ◮ – on the other hand, if we originally only had A j , and Page 4 of 12 then we add A i , then we switch our choice to A i . Go Back • Similarly, in the second case u ( A j ) > u ( A i ). Full Screen Close
Traditional Approach . . . A Testable . . . 4. The Above Testable Consequence is in Perfect The Above Testable . . . Agreement with Common Sense For Close Alternatives, . . . • Claim: the above behavior is in perfect agreement with Maybe Human . . . common sense. How to Take Into . . . Another Case when . . . • Case 1: the alternative A i is preferable to the alterna- Asymmetric . . . tive A j . How Does Our . . . • Expected behavior: choose A i irrespective of whether Potential Applications . . . we started with only A i or only A j . Home Page • Case 2: the alternative A j is preferable to the alterna- Title Page tive A i . ◭◭ ◮◮ • Expected behavior: choose A j irrespective of whether ◭ ◮ we started with only A i or only A j . Page 5 of 12 Go Back Full Screen Close
Traditional Approach . . . A Testable . . . 5. For Close Alternatives, Decision Makers Do Not The Above Testable . . . Behave in This Rational Fashion For Close Alternatives, . . . • Empirical result: when the alternatives are close in Maybe Human . . . value, decision maker exhibit “inertia”. How to Take Into . . . Another Case when . . . • Example: selecting between two similar retirement plans Asymmetric . . . A i and A j . How Does Our . . . • Case 1: we start with the plan A i and then add A j . Potential Applications . . . • Typical behavior: stick to A i . Home Page • Case 2: we start with the plan A j and then add A i . Title Page • Typical behavior: stick to A j . ◭◭ ◮◮ ◭ ◮ • Why this is counter-intuitive: Page 6 of 12 – if A i is better, then in Case 2, people should switch to A i ; Go Back – if A j is better, then in Case 1, people should switch Full Screen to A j . Close
Traditional Approach . . . A Testable . . . 6. Maybe Human Behavior Is Irrational? The Above Testable . . . • How can we explain this seemingly irrational behavior? For Close Alternatives, . . . Maybe Human . . . • One possible explanation is that many people do often How to Take Into . . . make bad (irrational) decisions: Another Case when . . . – waste money on gambling, Asymmetric . . . – waste one’s health or alcohol and drugs, etc. How Does Our . . . Potential Applications . . . • However, the above inertial behavior occurs among the Home Page most successful (otherwise rational) people. Title Page • It is therefore reasonable to look for an explanation of this seemingly irrational behavior. ◭◭ ◮◮ ◭ ◮ • It turns out that Page 7 of 12 – we can come up with such an explanation Go Back – if we take into account uncertainty related to deci- sion making. Full Screen Close
Traditional Approach . . . A Testable . . . 7. How to Take Into Account Uncertainty in De- The Above Testable . . . cision Making Situations For Close Alternatives, . . . • In practice, we can predict the consequences of alter- Maybe Human . . . natives only approximately, with some accuracy ε . How to Take Into . . . Another Case when . . . • So, instead of the exact values u ( A i ) and u ( A j ), we only know approximate values � u i and � u j . Asymmetric . . . How Does Our . . . • The actual utility values can be within intervals Potential Applications . . . u i = [ � u i − ε, � u i + ε ] and u j = [ � u j − ε, � u j + ε ]. Home Page • If the estimates are close, i.e., if | � u i − � u j | < 2 ε , then Title Page – there exist values u i ∈ u i and u j ∈ u j s.t. u i < u j ; ◭◭ ◮◮ and ◭ ◮ – there exist values u i ∈ u i and u j ∈ u j s.t. u i > u j . Page 8 of 12 • Thus, switching may decrease utility. Go Back • So, it is prudent not to switch (especially since often switching comes with a penalty). Full Screen Close
Traditional Approach . . . A Testable . . . 8. Another Case when Inertia is Beneficial: Con- The Above Testable . . . trol of a Mobile Robot For Close Alternatives, . . . • We change direction based on the moment-by-moment Maybe Human . . . measurements of the robot’s location and/or velocity. How to Take Into . . . Another Case when . . . • Measurements are never 100% accurate. Asymmetric . . . • The resulting measurement noise leads to random de- How Does Our . . . viations – shaking and “wobbling”. Potential Applications . . . • Each change in direction requires that energy from the Home Page robot’s battery go to the robot’s motor. Title Page • So, this wobbling drains the batteries and slows down ◭◭ ◮◮ the robot’s motion. ◭ ◮ • Natural idea: only change if it’s clear (beyond uncer- Page 9 of 12 tainty) that this will improve the performance. Go Back • Result: UTEP robot’s 1st place at 1997 AAAI compe- Full Screen tition. Close
Traditional Approach . . . A Testable . . . 9. Asymmetric Paternalism: Practical Applica- The Above Testable . . . tion of Present-Biased Preferences For Close Alternatives, . . . • Fact: the decision-making inertia is used in practice, Maybe Human . . . to encourage desirable behavior. How to Take Into . . . Another Case when . . . • Example: a kid can drink either a healthy fruit juice or a soda drink which has no health value. Asymmetric . . . How Does Our . . . • Traditional paternalism: prohibit undesirable choices. Potential Applications . . . • Problem: this enforcement rarely works. Home Page • More efficient idea: Title Page – at first provide only the desired alternative, ◭◭ ◮◮ – and then introduce all the other alternatives. ◭ ◮ • Example: have only healthy drinks for the first few Page 10 of 12 weeks of school, but then allow all the choices. Go Back • Result: due to inertia, kids tend to stick to their origi- Full Screen nal healthier choice. Close
Recommend
More recommend
