Searching for Privacy in all the Wrong Places: A behavioral economics perspective on individual concern for privacy George Loewenstein WEIS, 2007 Ed Lazear, "Economic Imperialism" Economic Journal,2000. • ECONOMICS • political science • history • demography • law • ….. "individuals engage in maximizing rational behavior" • • • • • • • • • • • • • spending/saving insurance labor market behavior bargaining investing medical decision making gambling criminal behavior fertility/sex charity/altruism dieting addiction suicide Behavioral Economics • ECONOMICS • psychology • neuroscience • medicine • law • ….. Behavior is often not ‘rational’ • • • • • • • • • • • • • spending/saving insurance labor market behavior bargaining investing medical decision making gambling criminal behavior fertility/sex charity/altruism dieting addiction suicide Behavioral economics is also imperialistic... • • • • • • behavioral finance behavioral labor economics behavioral public finance behavioral macroeconomics behavioral law and economics behavioral economics of information security? Focus of this talk: behavioral economics applied to understanding individual concerns/ behavioral responses to issues of privacy (neo)classical model of privacy Should I mention my sexual preferences on Facebook? (neo)classical model of privacy Maybe I’ll find a lover... But what about my future job prospects? And what if my parents happen to log on... (neo)classical model of privacy 1 1 p u ( benefits ) q i (1d )t i (1d )t u(cos tsit ) it $s WTA WTP Privacy Will discuss some complications... 1. Adaptation and loss aversion 2. Hyperbolic time-discounting 3. Preference uncertainty/constructed preferences ... then outline a simple perspective on individual privacy and present experimental tests of a few of its implications 1. Adaptation and loss aversion • Adaptation: People become accustomed to diverse circumstances – ownership – wealth – disabilities • Loss aversion: People dislike losing things relative to their present circumstances, but are often relatively indifferent to gaining those same things (loss aversion) • Also, people fail to predict these effects Evidence of adaptation: health conditions • Brickman, Coates, and Janoff-Bulman (1978) Surprisingly small difference in self-reported happiness (on 5 point scale) between paraplegics and matched controls: – paraplegics 2.96 – controls 3.82 • Wortman and Silver (1987): quadriplegics reported no greater frequency of negative affect than control respondents! • Tyc (1992): “no difference in quality of life or psychiatric symptomatology” in young patients who had lost limbs to cancer compared with those who had not Another manifestation of adaptation and loss aversion: the endowment effect Value don’t particularly like to gain Departures from reference point hate to lose Kahneman & Tversky (1979) An illustrative study • College student subjects • 6 attractive objects (e.g., ipod Nano, noise cancelling headphones) • 18 prices from 5% to 95% of retail price • Buy, sell, and choose trials • 2 of 6 products randomly assigned to each condition (buy, sell, choose) for each subject. • Subjects given two “sell” items to keep, and $20 • One of each type of trial counts Sell iPod Nano? $50 Buy Digital Camera? $40 Choose Wireless Mouse? $45 (get product) (get money) Choose Money? $55 (get money) (get product) Mean prices by condition $40 $30 $20 $10 $0 Buy Choose Sell People don’t anticipate adaptation (illustrated with the endowment effect) • Subjects: 27 CMU undergraduates & 39 Pitt MBAs • Procedure: 1. All shown mug 2. Half predict how much they would sell it for 3. All given a mug and opportunity to sell ________________________________________________________ Predicted and Actual Valuation of the mug n of predicted actual Group Condition subjects value value CMU Prediction 14 $3.73 $5.40 (0.41) (0.65) PITT No prediction 13 ------ $6.46 (0.54) Prediction 22 $3.27 (0.48) $4.56 (0.59) No prediction 17 ------ $4.98 (.53) (std. errors in parentheses) Implications for privacy • People will initially oppose losses of privacy • After loss of privacy, however, they will rapidly adapt • People will not be very motivated to gain new forms of privacy $s WTA WTP Privacy Concernwarming as a function as of problem severity of problem severity Concern about global a function (theoretical analysis) 24 2 Objective measure of severity 20 Subjective level of concern 16 12 1 8 0.5 4 0 0 0 5 10 15 20 25 time 30 35 40 45 50 Subjective level of concern Objective measure of severity 1.5 Concern aboutConcern globalaswarming as a function of problem severity a function of problem severity (theoretical analysis) 2 30 Objective measure of problem (no policy) 25 Objective measure of problem (with policy) 20 Subjective level of concern (no policy) 1.5 1 15 10 0.5 5 0 0 0 5 10 15 20 25 time 30 35 40 45 50 Subjective level of concern Objective measure of problem 35 2. Time-discounting • Ideal: people balance present and future costs & benefits in an even-handed fashion • Reality: people place disproportionate weight on the present, relative to all future periods; ‘hyperbolic time discounting’ • Especially true of: – young people – people who are in emotional states – people who are distracted 'Schindler's List' study In exchange for completing an unrelated survey, subjects receive three free video loans from our collection Our collection... Lowbrow movies The Breakfast Club (1985; 2.3) Clear and Present Danger (1994; 2.8) Four Weddings and a Funeral (1993; 3.1) Groundhog Day (1993; 2.4) I Love Trouble (1994; 2.0) In the Line of Fire (1993; 3.0) Indecent Proposal (1994; 2.9) The Mask (1994; 1.6) Mrs Doubtfire (1993; 2.9) My Cousin Vinny (1992; 3.0) Sleepless in Seattle (1993; 2.5) So I married an Axe Murderer (1993; 2.3) The Specialist (1994; 2.1) Speed (1994; 2.5) Highbrow movies Blue (1993; 5.1) (subtitled) Blue Sky (1994; 3.8) (Oscar winner) Dear Diary (1994; 3.1) (subtitled) Hoop Dreams (1993; 3.3) (documentary) Like Water for Chocolate (1993; 4.5) (subtitled) Naked (1993; 4.0) Raise the Red Lantern (1991; 3.7) (subtitled) Schindler's List (1993; 6.8) (Oscar winner) The Piano (1993; 5.7) (Oscar winner) The Scent of Green Papaya (1993; 3.8) (subtitled) Conditions: • sequential: Subjects choose each movie on the day when they will watch it • simultaneous: Subjects choose all three movies on the first day (when they will watch the first movie) Results: Percent of highbrow movies chosen for each day of choice Experiment 1 Day of Choice Condition One Two Three Sequential 42% 47% 44% Simultaneous 44% 63% 71% Schindler's list chosen only once to be watched on the day of choice; 13 times on future days. Implications for privacy: People won’t weigh short-term benefits of divulgence against long-term consequences for privacy in evenhanded fashion Intertemporal choice as a metaphor • Early view: single discount rate for individual applies to all forms of consumption. Then... – Steeper discounting for shorter time delays (hyperbolic t.d.) – Steeper discounting for small outcomes than large outcomes (Thaler, 1981) – Steeper discounting for gains than for losses (Thaler, 1981) – Steeper discounting for delay than for speed-up (Loewenstein, 1988) – Negative discounting for sequences (e.g., Loewenstein & Prelec, 1993) Present ‘state of the art’: multiple motives (Frederick & Loewenstein, under review) Choice Pricing Allocation THIS MONTH NEXT MONTH A: $50 dinner ........ $100 dinner B: $100 dinner ........ $50 dinner A: $50 dinner ........ $100 dinner B: $100 dinner ........ $50 dinner $75 $75 53% 47% $67 $78 Privacy similar to intertemporal choice... • Multiple motives in decisions involving privacy – Encouraging divulgence: • material benefits to exchange of information • desire to be known (e.g., myspace, facebook, posting of personal pictures), for fame (or even notoriety) • desire to be part of a group – Discouraging divulgence: • material consequences • innate/inchoate qualms about revealing information • Which motives dominate likely to depend on subtle factors, e.g., – medium (email, web, phone, face-to-face) – Warnings • Often what matters is not what should matter – e.g., email versus phone 3. Preference uncertainty & constructed preferences • People don’t know what they want or what they care about • However, people often respond sensibly to changes in their environment ‘coherent arbitrariness’ Illustrative Study (Experiment 4, from Ariely, Loewenstein & Prelec, 2003) • Stimuli: Unpleasant noises played over headphones • Anchor manipulation. Hypothetically.. "would you listen to this noise for 300 seconds for $x.00?" (anchor based on first three digits of subject’s social security number – e.g., 478 = $4.78) • Subjects listen to sample noise for 300 seconds • Subjects state WTA for 3 noises that differ in duration: increasing condition: 100 sec, 300 sec, 600 secs. decreasing condition: 600 sec, 300 sec, 100 secs. • Truthful elicitation procedure; if WTA<randomly drawn price, then they hear the tone for that price Results • Coherent arbitrariness can be seen in numerous domains, e.g., –labor supply –criminal deterrence –financial markets (efficient markets) Implication for privacy: people don’t have a clue about how important privacy is; however, they are likely to respond sensibly to changes Putting the pieces together; A behavioral perspective on concern for privacy: • People generally not concerned about privacy • Many other motives are more powerful – e.g., desire to be known, desire for fame • Only concerned about privacy if alerted that privacy could be or is being violated and could have consequences – Cues: • Warnings • Prying • Explicit comparisons (of privacy to no privacy) Two studies of privacy (collaborations with Alessandro Acquisti and Leslie John) Study 1: paradoxical effects of reassurance Thesis: If people don’t naturally think about privacy, then reassuring them can potentially cause them to be more rather than less cautious about divulging information Design • Survey respondents asked for email, then asked 14 questions, 6 about sensitive information • Three conditions: – No privacy/anonymity assurance – Weak assurance “A quick note to let you know that any identifying information you may choose to provide in this survey will be stored separately from your responses. In addition, your survey responses will only be analyzed in aggregate.” – Strong assurance “Concerning the confidentiality and anonymity of your responses: Please be advised that maintaining the confidentiality and anonymity of your responses is of the utmost importance to us. The following procedure will be used to maintain your anonymity in analysis, publication, and presentation of any results. Anonymity will be maintained during data analysis and publication/presentation of results by any or all of the following means: (1) You will be assigned a number as names will not be recorded. (2) The researchers will save the data file by your number, not by name. (3) Only members of the research group will view collected data in detail. (4) Any recordings or files will be stored in a secured location accessed only by authorized researchers.” Please answer the following questions, which refer to your educational experience since high school. Yes No 1. Since high school, have you ever handed an assignment in late? 2. Are you currently taking at least four courses? 3. Have you ever plagiarized text for any kind of assignment? 4. Have you ever let a classmate copy from you during an exam? 5. Do you arrive late to class more often than the majority of your classmates? 6. On average, do you find the number of students in your classes to be conducive to learning? 7. Have you ever failed a course? 8. Have you ever copied a classmate’s homework? 9. What is your grade point average? 10. Have you ever cheated on an exam? 11. Have you ever requested an extension for an assignment? 12. Do you regularly attend classes? 13. Have you ever lied to a teacher in order to avoid taking an exam or handing in a term paper on time? 14. Have you ever lied about your grade point average? GPA: ____________ results 1. Most people would tell us almost anything and everything (even after giving us an email address that could easily be used to identify them) 2. Small reassurance had little effect, but substantive reassurance backfired (as predicted) Innocuous Questions (6) Probability of responding affirmatively 0.6 0.5 0.4 0.3 0.2 0.1 0 no assurance weak assurance strong assurance Probability of responding affirmatively Sensitive Questions (6) 0.25 0.2 0.15 0.1 0.05 0 no assurance weak assurance strong assurance Study 2: explicit versus implicit inquiries about sensitive behaviors Initial page from which NYTs readers are diverted to different survey versions... Three conditions.. • Baseline: simply asks respondent whether or not they have ever engaged in 34 different behaviors ranging from very mild (e.g., leaving lights turned on) to very severe (e.g., cheating on taxes) • Commission: asks respondents to rate how unethical the 34 activities are, but only if they have engaged in them • Omission: asks respondents to rate how unethical the 34 activities are, but only if they have not engaged in them commission condition... commission condition... Results.. Have you ever engaged in behavior (yes/no)? (n=241) Commission (answer only if you have EVER engaged in behavior) (n=237) Omission (answer only if you have NEVER engaged in behavior (n=213) 8.3% 16.6% 15.5% Having sex with someone who is too drunk 5.3% to know what they are doing 15.2% 13.5% Trying to gain access to someone else’s (e.g., a partner, friend or colleague’s) email account 31.5% 26.6% 33.8% Making a false insurance claim 5.8% 8.9% 22.1% Cheating on one’s tax return 15.8% 17.3% 28.2% Having sex with the current husband, wife or partner of a friend 9.6% 13.9% 25% Mean 12.7% 16.4% 23.0% Stealing something worth more than $100 Conclusions • Most people don’t inherently care much about privacy • Individual intuitions and emotions provide a poor guide to privacy-related behavior • New technologies... – greatly magnify the risks – e.g., internet posting cannot be undone – eliminate danger cues -- e.g., email flaming – introduce new cues (e.g., indications of community) that actually mute privacy concern – e.g., Facebook need for new regulation or greater tolerance for individual idiosyncrasies