What is risk taking? In their seminal work, Jessor andJessor (1977) de?ned risk taking as “behavior that is socially de?ned as aproblem, a source of concern, or as undesirable by the norms of conventionalsociety and the institutions of adult authority, and its occurrence usuallyelicits some kind of social control response” (Jessor & Jessor 1977, 33.o.).While the lay and clinical definition of risk-takingis often used in the sense of engaging in a behavior that could potentiallyhave a negative outcome.
(Defoe et al., 2015). There are many validated and frequently used risk-taking tasks that we can choose from when planning an experiment. Despite thenumerous refined risky decision-making tasks, there is still no consensus aboutthe definition of risk. Decision theory usuallydescribes risky situations as choices between lotteries characterized byoutputs (gains and/or losses) and their probabilities (Defoe et al., 2015). The core characteristic of the term risk as used inthe judgment and decision literature is outcome variability: The option withthe widest range of possible outcomes is considered the riskiest option (Defoe et al., 2015).
All these tasks measure the risk one is willing totake in a situation, but which of them is able to predict real life risktaking? Do they measure the same thing or are there differences? That is thequestion I would like to answer in this paper. Studies show that cognition isdifficult and costy, partially because of peoples’ limited processing capacity,including attention (Markiewicz& Kubi?ska, 2015). Among other things, this is why decision makers firsttry to simplify the problem in a so called “editing phase” (Markiewicz & Kubi?ska, 2015). Because ofattention limitations they do not use all of the available information whenmaking a decision. Although many researce investigate the order of information,relatively little has been said about differences in information use inaffective (hot) and cognitive (cold) risk processing. Do people concentrate ondifferent risk characteristics (losses of, and gains on, stakes, and theirprobabilities) in emotional risk taking (e.
g., parachute jumping) compared tocognitive risk taking (e.g., pension scheme decisions)? Some studies havedemonstrated that the impact of probabilities is strongly diminished foraffect-rich outcomes (Markiewicz& Kubi?ska, 2015). However, these studies used outcomes of differentvalences, assuming that medical outcomes are affect-rich and that monetaryoutcomes are affect-poor (Markiewicz& Kubi?ska, 2015).
Most ofthe tests listed below are measuring affective (hot) risk taking. BallonAnalogue Risk Taking-Youth versionThe BART-Y test is a version of theBART test for adolencents. The participants (pubescentadolescents) are instructed to pump a balloon.
The explosion point ofthe ballon is randomized and varies per every trial. With each pump theparticipant can win one point, but each pump also increases the chance of anexplosion, resulting the loss of all the gained points for that balloon. Ifparticipants stop pumping the balloon, they then earn all of the points accumulatedso far for that. Risk-takingwas measured as the average number of pumps on unexploded balloons.
The BART-Y task is considered to be a hot risk takingtask, as the reward or loss is presented right after the participant makes thedecision about stopping or continuing the pumping. Although, the strength ofthe influence of the emotions can be reduced by delaying the rewards, whichcould make this task a little less like a hot task. BART score is related to selfreported engagement inreal-world risk-taking behaviors including substance use and delinquency/safetybehaviors from middle adolescents to young adults (Dahne et al., 2013). Thetask was found to be associated with self-reported risk behaviors, such asalcohol use, substance use, gambling, delinquency, and risky sexual behavior (Dahneet al., 2013). BART performance appears to be related to early engagement insubstance use as well as to risk behaviors that relates to substance use andalso the task seem to predict young adolescents’ propensity for future riskseven if they have not yet started to engage in health-compromising activities (Dahneet al.
, 2013). The wheel of fortuneThe computerized version of the wheel of fortune (WOF)task is a 2-choice decision-making task with probabilistic monetary outcomes (Defoe et al., 2015). On each trial, a wheel (a circle divided into 2 slices of different sizeand of 2 different colors) is presented to participants (82 adolescents with nolifetime history of psychiatric illness). Throughout the task, 4 types ofmonetary wheels are presented in random order, differing on probability of thereward magnitude.
Smaller slices are always paired with the higher reward magnitude. Participants are instructed to select 1 ofthe slices, naming the color. If the computer randomly selects the same coloras the participant does, the participant wins the designated amount of money.However, the participant wins nothing if the computer randomly selects theother color. Risk-taking was measured with a percent risky selections score,which was computed using the number of times 10% and 30% probability optionswere selected relative to the total number of times that the 10/90 and 30/70wheels were presented (Defoe etal.
, 2015). Participants got feedback right after they made thedecision and chose a wheel; therefore this task is also considered as a hotrisk taking task.Greater frequency of low-probability (high-risk)choices on the win–no win version of the WOF predicted substance-relatedproblems, including drug involvement and related psychiatric and psychosocialproblems (Dahne et al.
, 2013). However, lowprobability (low-risk) choice on thelose –no lose version of the task did not predict substance-related problems(Dahne et al., 2013). Although individual differences in risky selections onthe WOF were associated with risk-taking behavior and substance relatedproblems, the prevalence of substance-related problems was low, probablybecause this was a selected group at extremely low-risk for psychopathology.
The Iowa Gambling TaskThis task is quite similar to the previous one. Onboth, participants (substance abusers and non-substance abusers of cocaine andmarijuana) have to choose between high-risk, high gain and low-risk, low gainoptions. In this task for each trial, participants have to choose one card at atime from 1 of 4 decks that differ in payoffs and losses.
Selections from the 2″disadvantageous” decks are followed by a higher reward (on most trials), butalso by higher (unpredictable) losses; thus, the final result is “overall netloss” (Defoe et al., 2015). The 2 “advantageous” decks are followed by lowerrewards on most trials but also by lower (unpredictable) losses; thus, the finalresult is “overall net gain”. Risk-taking wasoperationalized as the mean numberof choices from the deck with the highest outcome variability ( the “risky”deck).
Participants learn about the experienced outcomes and the differencesbetween the decks through trials and errors, as the feedback about thedecisions they made are immediate, which also make this task a hot risk takingtask. In studies using the IGT, risk-taking is typically operationalized as thenumber of choices from the 2 advantageous decks minus the 2 disadvantageousdecks (Defoe et al.,2015). Results indicated that both cocaine users andmarijuana users performed worse than controls on the total IGT net score (totalscore across sessions 1 and 2) (Dahne et al.
, 2013). Furthermore, all groupsexhibited between- session learning, but the rate of learning differed betweengroups such that cocaine users exhibited less learning than marijuana users andmarijuana users exhibited less learning than control (Dahne et al., 2013). The framing spinner taskIn the Framing Spinner Task, participants (153students) make a choice between 2 spinners with an arrow in the middle: One spinneris completely red representing a sure option and the other spinner had varyingproportions of blue and red representing a gamble. Risk levels varied asfollows: one-half, two-thirds, and three-fourths chance of winning nothing (gainframe) and one-half, two-thirds, and three-fourths chance of losing something (lossframe) (Reyna, Estrada, DeMarinis,Myers, Stanisz, & Mills, 2011).
Reward levels varied between low ($5), medium ($20), and high ($150). Inloss problems, participants began with an endowment, from which subsequentlosses were deducted, whereas in the gain frames participants begin with nomoney. The displayed net outcomes were the same for both frames. Risk-takingwas operationalized as the proportion of gamble choices. In this hot risktaking task, on each trial, after participants selected their choice, they ratedtheir degree of preference too, which only strenghtened their feelings aboutmaking the decision and made them concentrate more on the emotional part oftheir decision making. In this task reasoning was the most consistent predictorof real-life risk taking. The Gist factor (The four gist measures included theCategorical Risk scale, the Gist Principles scale, a Global Risk question, anda Global Benefits question) was associated with fewer sexual partners, but theVerbatim/Reverse Framing factor was associated with more sexual partners (Reyna et al., 2011).
Intentionsto have sex, sexual behavior, and number of partners decreased when gist-basedreasoning was triggered by retrieval cues in questions about perceived risk,whereas intentions to have sex and number of partners increased whenverbatim-based reasoning was triggered by different retrieval cues in questionsabout perceived risk (Reyna et al.,2011). The knife switches task (Thedaredevil task)The participants (44 children ageing from 4 years 9months to 6 years 5 months; mean age: 5 years 6 months) are seated in front ofa panel of 10 small knife switches and are told that 9 of these switches were”safe” and one was a “disaster” switch.
The participant is instructed to pullone of the switches. If the participant pulls a safe switch, he (or she) isallowed to put one spoon full of M&M’s candies into a glass bowl. Theparticipant then has to decide whether to pull another switch in an attempt towin another spoonful of candy or to stop and keep the candy.
If a participantpulls the disaster switch, he (or she) looses all the accumulated candy. Thegame ends when the participant either stops and collects his candy or pulls thedisaster switch and loose all of it. Risk-taking was operationalized as thenumber of pulled switched.
This task is also a hot risk taking task, asparticipants can see the tempting reward (the candy) and got the feedback aboutgaining or not gaining some immediately after making the decision. In the research, participants were devided into twogroups (risk avoiding and risk-taking) according to their performance in theknife switches task, and then they had to participate in a traffic task (Hoffrage, Weber, Hertwig & Chase, 2003). In thetask a car heading towards the point the children were supposed to cross thestreet was presented and the children had to make a stop (not crossing thestreet) or a go (crossing the street) decision. The researchers measured themaximum time and distance in which the participant is still willing to crossthe street in front of the car. According to the research, risk takers had asignificantly higher hypothetical accident rate (3.7%; 61 of 1,654) than riskavoiders (0.6%; 5 of 857) (Hoffrage, Weber,Hertwig & Chase, 2003). This difference appears to be small, however,when viewed in terms of Rosenthal and Rubin’s (1982) binomial effect-sizedisplay, it amounts to a difference of 9 percentage points between the twogroups (Hoffrage, Weber, Hertwig & Chase,2003).
Risk takers (using thegambling classification) tolerated shorter leeway times than risk avoiders and alsohad a higher hypothetical accident rate, that is, a higher percentage of godecisions (out of all gaps) that left a leeway time of less than 3 s –which is thetime children needed to run across the street on average (Hoffrage, Weber, Hertwig & Chase, 2003). ColumbiaCardsorting Task (Hot version; Cold version)The CCT begins with a presentation of32 (gain or loss) cards and a score of 0 points. Participants (497 students) are askedto turn over cards. A round ends when participants encounter a loss card, or ifthe participants stop turning over cards to collect all the gains. Per round,three variables vary systematically: the magnitude of gain, the magnitude ofloss, and the gain/loss probability (Defoe et al.
, 2015).In the Cold version of the CCT,participants state in advance how many cards they want to turn over, in the hotversion participants turn over cards one-by-one until they decide to stop. Inthe Hot version, participants receive feedback immediately after turning over acard, while in the Cold version they receive feedback at the end of the finalround. On average, respondents performing the Hottask disclosed more cards (M = 27.185; SD =3.
173) than those taking part in the Cold condition (M =14.000; SD = 5.303) (Markiewicz & Kubi?ska, 2015). Thoseparticipating in the cold condition also displayed higher information use, regardlessof the measure employed. Cold condition participants paid more attention(compared to hot condition) to the amount of gain and less to probabilityinformation (Markiewicz & Kubi?ska,2015).Our results are in line with other CCT studies (Buelow, 2015) saying “It appears thatparticipants in the CCT-cold condition paid greater attention to all threeinformation use variables in making decisions than those in the CCT-hotcondition, in which riskier performance was only associated with lossprobability.” (Markiewicz & Kubi?ska, 2015,184.
o.). DiscussionIn this paper I examined 6 risk taking tasks (one coldand 4 hot task), helping reasearchers to decide in what context which taskshould be used. Altough relatively little can we tell about cold risk taking tasks, asthere are a lot less task that measures the cold aspect of risk taking, thanones measuring the hot aspect, it seems like decision makers place differentweights on risky situation parameters (gain amount, loss amount, andprobability of gains), while making risky decisions in cold and hot tasks.
I have found that the above mentioned tasks werevalidated in very different ways. Some were correlated with questionaires andsome with risk taking in traffic situations. Some are correlated withexperiences that actually happened in the past or with the subject’s currentlifestyle (BART-Y, Wheel of fortune, IGT) and some are correlated withimaginary/ hypothetical situations in the present (The knife switches task). Asthese aspects are very different, they might do not measure the same kind ofrisk taking.
Can we expect correlation between cold and hot tasks? Also it is questionable, if we can announcethat the test measures risk taking, just because it is validated by anothertype of risk taking task. It is also a good question, whether it is importantto correlate with other tests, as they most likely will not measure exactly thesame. To understand thesetests more and find answers to these questions, a metaanalysis should be done,correlating all the risk taking tasks with each other.
The conclusion is, thatthere are many aspects of risk taking that can be measured and we cannot claimthat one is better than the other.