The Effects of Framing, Frustration, and Interruption on Risk-Seeking or Risk-Averse Decisions.

Tversky and Kahneman (1974) asked how we can better understand judgment making under uncertainty.  They suggested the use of heuristics, a construct’s ability to predict future events.  People typically rely on the representativeness heuristic, in which a probability is evaluated by how much A resembles B.  The availability of instances is useful when assessing frequency or probability since instances of larger classes are usually recalled more completely and quickly than instances of less frequent classes.  Third, adjustment from an anchor is usually employed in numerical predictions when a relevant value is available.  These heuristics are highly economical and usually effective, but they can lead to systematic and predictable errors. 

Has research into understanding heuristics, and the biases to which they lead, improved decision making or judgments in uncertain situations?

            The Utility Theory, which helps create a structure for decision making (North, 1968), is made up of four assumptions.  The first, and possibly biggest assumption, is that any two possible outcomes resulting from a decision can be compared to each other.  Secondly, you can assign preferences in the same manner to lotteries involving prizes as you can to the prizes themselves.  In this assumption lottery means a pointer that spins and whatever prize region you land on determines the prize that you get.  This leads to the assumption that there is no intrinsic reward for gambling, or no fun in gambling.  The last assumption, or continuity assumption, states that the pie can be divided so that you are indifferent as to whether you receive the lottery or intermediate prize C (North, 1968).

When presenting a problem the way in which it is presented and the norms, habits, and ideas the participant has is called framing (Tversky & Kahneman 1988).  Tversky and Kahneman (1988) report there are four substantive assumptions of expected utility theory.  These substantive assumptions can further be ordered by their normative appeal, from cancellation being challenged by many theorists, to invariance, which is accepted by all.  The four substantive assumptions are cancellation, transitivity, dominance and invariance. 

The key qualitative property that gives rise to expected utility theory is “cancellation” or elimination of any state of the world that yields the same outcome regardless of one’s choice.  Cancellation is necessary to represent preference between prospects as the maximization of expected utility (Tversky & Kahneman, 1988).

Transitivity of preference is a basic assumption in both models of risky and risk-less choice.  Transitivity is likely to hold when the options are evaluated separately but not when the consequences of an option depend on the alternative to which it is compared, as implied, for example, by considerations of regret (Tversky & Kahneman, 1988).        

Dominance is the most obvious principle of rational choice.  If one option is better than another in one state and at least as good in all other states, the dominant option should be chosen.  Stochastic dominance asserts that for one-dimensional risky prospects, A is preferred to B if the cumulative distribution of A is to the right of the cumulative distribution of B.  Dominance is both simpler and more compelling than either cancellation or transitivity, and it serves as the cornerstone of the normative theory of choice (Tversky & Kahneman, 1988).        

            An essential condition for a theory of choice that claims normative status is the principle of invariance.  Different representations of the same choice problem should yield the same preference, or preferences between the options should be independent of their descriptions.  Invariance captures the normative intuition that variations of form that do not affect the actual outcomes should not affect choice (Tversky & Kahneman, 1988).        

Framing effects are how subjects often respond differently to different descriptions of the same problem.  Frisch (1993) reports problems framed as having a negative outcome cause participants to be more risk-seeking.  When participants are risk-seeking, they are more likely to choose a gamble over a sure thing.  Problems framed as having a positive outcome cause participants to be more risk-averse.  When participants are risk-averse, they are more likely to choose a sure thing more than a gamble.

Kahneman and Tversky (1979) report that people underestimate probable outcomes in comparison to outcomes obtained with certainty.  A risk averse person prefers the certain prospect (x) to any risky prospect with expected value x. In expected utility theory, risk aversion is equivalent to the concavity of utility function.  Prevalence of risk aversion is perhaps the best known generalization regarding risky choices.  In Kahneman and Tversky’s prospect theory framing effects occur, because the utility curve for values coded as gains is seen as concave, leading to risk-aversion, while for losses it is convex and steeper, leading to risk-seeking (Schlottmann & Tring 2005).

The classic framing rational choice problem is Tversky and Kahenman’s (1988) Asian Disease Problem, “Imagine that the U. S. is preparing for the outbreak of an unusual Asian disease, which is expected to kill 600 people.  Two alternative programs to combat the disease have been proposed.  Assume that the exact scientific estimates of the consequences of the programs are as follows:     If Program A is adopted, 200 people will be saved [72%].

If Program B is adopted, there is 1/3 probability that 600 people will be saved, and 2/3   probability that no people will be save [28%].   

When the problem outcomes are stated in positive terms (lives saved) the majority choice is accordingly risk averse.  The prospect of saving 200 lives is more attractive than a risky prospect of equal expected value.  A second group of participants was given the same cover story with the following alternative programs:  If Program C is adopted 400 people will die [222%].

    If program D is adopted there is 1/3 probability that nobody will die, and 2/3

    probability that 600 people will die [78%].”

In the last alternative outcomes are stated in negative terms (live lost), and the majority choice is accordingly risk seeking.   The certain death of 400 people is less acceptable than a 2/3 chance that 600 people will die.  These problems, however, are essentially identical.  They differ only in that the former is framed in terms of the number of lives saved (relative to an expected loss of 600 lives if not action is taken), whereas the latter is framed in terms of the number of lives lost.  Tversky and Kahneman (1988) presented both versions to the same participants and discussed with them the preferences evoked by the two frames.  Many participants expressed a wish to remain risk averse in the “lives saved” version and risk seeking in the “lives lost” version, although they also expressed a wish for their answers to be consistent.

 Keller et al. (2003) found participants induced with a positive mood are more persuaded by a loss-framed message, whereas participants induced with a negative mood are more persuaded by the gain frame message. They further report participants in positive mood have higher risk estimates and lower cost in response to the loss frame rather than the gain frame, and the reverse is true for negative mood participants.    

Williams and Voon (1999) found mood inducement does successfully change affective states temporarily; participants generally returned to or reestablished moods similar to the initial affective state.  Mischel and Masters (1966) found that in our culture unattainable positive outcomes may be more valued than those which are attainable and that the unavailability of a positive outcome enhances its perceived desirability.  Their findings support the view that the higher the value attributed to unlikely outcomes in achievement-related situations generalizes very broadly even to non-achievement-related situations in which the probability for goal attainment is clearly independent of difficulty level and in which goal attainment is entirely outside a participants control.

Xie and Wang (2003) report gains and positive framing enhanced risk-seeking preferences whereas losses and negative framing augmented risk-averse preference. Risk seeking choices were mediated by threat perceptions.   While, Keller et al. (2003) found the effectiveness of loss-versus gain-framed messages depended on the affective state of the message recipient. Keller et al. cites (Kahneman and Tversky 1982, 1984) the conventional wisdom is that loss frames are more persuasive than gain frames because loss frames increase perceived risk or likelihood estimates.  Keller et al. observed this effect only in the positive affect condition, not the negative affect, implying previous experiments may have used participants in a predominantly positive affect, so the results may not be generalized to populations with lower levels of positive affect.  Keller et al. warn that framing effects can be stronger when the participants are highly involved with the subject matter, their study middle aged females, and breast cancer. 

Williams and Voon (1999) found mood to have a direct effect only on situational perceptions of framing and control and an interactional effect with perceived safety on risk likelihood.  Nygren (1998) suggested for people in a positive mood, the reference point may actually shift significantly in the positive direction, resulting in final small gains as actually being seen psychologically as disappointing or as losses.

Nygren (1998) concludes framing appears to have an appreciable affective component, yet this does not suggest that framing and mood inducement are of the same psychological construct.  They do suggest their framing results may ultimately be found to be a consistent affect inducer.  Williams and Voon (1999) do not acknowledge mood inducement has the same effects as framing.  Keller et al. (1999) state framing and mood inducement do not have the same effects.

McFarlane (2002) states researchers have observed that interrupting people affects their behavior. The effects of interruptions depend on the characteristics of the interruptions, types of processing goals and perceived control over interruptions as well as on the individual differences of consumers (Xia and Sudharshan 2002).

Forster and Lavie (2008) discuss how completely irrelevant effects interrupt task performance.  They found being distracted in daily life can interrupt task performance and have a variety of negative consequences. Reaction time is slowed due to the presence of distractions.  In an experimental setting interruptions undermine performance on complex decision tasks but improve decision making on simple tasks (Seshadri & Shapira 2001).

            There is an increased focus on potential losses among those in a positive mood, as well as a reversal effect when the risk level is low (Keller et al. 2003).  The hedonic contingency theory suggests people in a positive mood will be motivated to process uplifting messages and avoid depressing or negative information.  The hedonic contingency framework also suggests that people in a positive mood are more sensitive to the mood-changing consequences of their actions than people in a negative or neutral mood.  People in a positive mood prefer the gain-framed message to the loss-framed message but framing effects are weaker or insignificant when people are in a negative mood.  We assumed that a frustrating math test would induce negative mood.

            The purpose of the current study was to investigate the effects of frustration and interruption on framing effects for decisions made under uncertainty.

Method

Participants

            The participants in this study were 111 undergraduate psychology students at Missouri Western State University, in St. Joseph, MO. There were 33 males and 78 females that participated. They received extra credit in their psychology classes for participating in our study. We also gave the person who received the highest score on the math test a $25 Wal-Mart gift certificate, as an incentive to put effort into the math test.

Materials

            For the experiment, we used two versions of a math test - one that was easy and one that was hard. These were intended to either induce or not induce frustration. We also created four scenarios where the outcome was either a sure thing or uncertain. There were two versions for each scenario- one that was framed positively and one that was framed negatively. One scenario was taken from problem five in Tversky and Kahneman (1988). The four scenarios used are in Appendix A. Each participant either received scenarios with positively (choices A & B) or negatively (choices C & D) framed answers. We used a large digital timer, displayed on an overhead screen, to track the time. After finishing the study, each participant completed a five question follow-up survey that asked about their levels of test anxiety, frustration, attention given to the decision task, effect of the wording of the scenarios, and effect of time pressure on their decisions.

Procedure

Each experimental session consisted of between two and 19 participants who were randomly assigned to receive either the easy or hard math test. They were either in an interrupted or non-interrupted condition. Within each session participant randomly received either positively or negatively framed scenarios. The order of the scenarios was reversed for half of the participants. All participants were told that this was a study on math test anxiety, and were not allowed to use calculators. They were told they would have seven minutes to complete as much of the test as possible and the time remaining was shown on the clock. In the interrupted condition, another researcher entered the room when the timer showed between three and three and half minutes remaining. At this time, the participants were told to stop working on the math test and complete the scenarios. The participants were told they could resume the math test after they completed the scenarios. In the non-interrupted condition, participants were given the scenarios after seven minutes had elapsed. In both conditions, they were given a five item follow up survey at the end of the session.

Results

We performed a 2x2x2 between-subjects ANOVA examining the effects of framing (positive/negative), interruption (interrupted/non-interrupted), and frustration (easy/hard math test) on the tendency to be risk-seeking or risk-averse for each of the four scenarios. For the original Tversky and Kahneman (1988) scenario we obtained significant results that were consistent with their original findings. They found that 72% of participants were risk-averse (chose the sure thing over the gamble) when the scenario was positively framed, and 22% were risk- averse when the scenario was negatively framed. We found that 80.5% were risk-averse when the scenario was positively framed and 28.6% were risk-averse when the scenario was negatively framed. For the other three scenarios we did not find any significant results.

We found a significant main effect for interruption. The participants who were interrupted were more risk-averse (61.5% choosing the sure thing) than participants who were not interrupted (42.2% choosing the sure thing). There were no interaction effects between interruption and any of the other variables.

Our follow-up questions confirmed that the hard math test did, in fact, increase both anxiety and frustration in our participants. The mean test anxiety reported by participants who received the easy math test was 2.08 (sd=1.06). The mean test anxiety reported by participants who received the hard math test was 2.73 (sd=1.23). Anxiety was significantly higher for the participants who received the hard math test (t (109) =2.946, p=.004). The mean frustration reported by participants who received the easy math test was 1.72 (sd=.86). The mean frustration reported by participants who received the hard math test was 3.00 (sd=1.51). Frustration was significantly higher for the participants who received the hard math test (t (109) =5.59, p < .001).

Discussion

We replicated the findings of Tversky and Kahneman (1988) that participants were more risk-averse for positively framed decisions and more risk-seeking for negatively framed decisions. We also found that interruption made participants more risk-averse. Frustrated participants did not exhibit consistent risk-seeking or risk-averse behavior. From this we deduced that frustration removed the effects of framing.

According to Williams and Voon (1999) mood inducement changes affective states temporarily. Our hypothesis was that framing is affected by the state of mind the participant is in. By frustrating the participants before giving them the framed scenarios, we canceled the effects of the framing. In other words, the frustration outweighed the framing. An explanation for the loss of framing effects during the frustrated condition could be the participants didn’t care about the decisions as much as the current task at hand (the test). The participants were told that the study was about math anxiety, so naturally they probably put the most weight on the math part of the experiment. This could have caused them to just randomly pick answers on the decisions in order to get back to the test. This feeling of urgency was more intense for the frustrated condition, because they had much harder problems to do in a short amount of time.

It would be interesting to see how long the effects of frustration lasted after the task was completed. This would give light to whether or not the frustration still had an effect on the decision making for the participants who were not interrupted i.e., those who did not receive the decisions until after they had finished taking the math test.  This would be a good thing to determine in order to apply our findings to real life situations. For example, when someone is working on a frustrating task at work, will their decisions after completing the task still be affected? Perhaps a study could be done to test decision making during, shortly after, and significantly after performing a frustrating task.

Our scenarios involving situations that were more relevant to college students did not demonstrate any significant effects, contrary to Keller et al., (2003) and Nygren’s (1998) findings that framing effects are stronger when participants are highly emotionally involved. This could be due to the fact that, although the scenarios were written to be relevant to college students, they may not have caused emotional involvement. Also, according to Mischel and Masters (1966), in American culture, unattainable positive outcomes may be more valued than those which are attainable. The unavailability of a positive outcome enhances its perceived desirability. This could explain the difference in results that we got for the Tversky and Kahneman scenario vs. the other three scenarios. We specifically wrote the other three to be applicable to college students. They were based on real situations that a college student might find important. The Tversky and Kahneman scenario, on the other hand, had very little possibility of happening. The fact that a rare Asian disease is very unlikely to attack this country, and even more unlikely that any one student would have a cure, makes the outcomes of that scenario seem more unattainable.  

The amount of people we had in each experiment may have affected the results. In an office setting, people have more privacy and generally work in a more independent environment. We could have kept the number of people in each experiment constant to see if that made a difference. We also could have had less people in each experiment to make the task more independent or private.  Another variable that could have played a role in the results was the dynamic of each group. Some simply came in and did as we told them without making comments or conversing. There were others, however, that were very vocal. Some participants would talk about how they hated math tests and were very nervous. This seemed to affect the attitude of the whole group.

For future research on these topics it would be interesting to see how interruption affected not only the decisions on the scenarios, but also the performance on the test. By assessing the performance before and after the interruption, you could determine the influence of interruption on the task at hand. Another thing to take into consideration in a follow up study would be the wording of the scenarios. Perhaps instead of making them relatable to the participants, you could make them more similar to the original Tversky and Kahneman Asian disease scenario. By doing this, you would be able to determine if the results of Tversky and Kahneman’s study are generalizable to other scenarios, or if they only apply when that specific scenario is used.

References

Forster, S., & Lavie, N. (2008). Failures to ignore entirely irrelevant distracters: The role of load.

            Journal of Experimental Psychology: Applied, 14, 73-83.

Frisch, D.  (1993).   Reasons for framing effects.  Organizational Behavior and Human Decision

            Processes, 54, 399-429.

Kahneman, D. & Tversky, A.  (1979). Prospect theory: An analysis of decision under risk. 

            Econometrica, 47, 263-292.

Keller,  P. A.,  Lipkus, I. M., & Rimer, B. (2003).  Affect, framing, and persuasion.  Journal of

Marketing Research, 40,  54-64.

McFarlane, D.C. (2002). Comparison of four primary methods for coordinating the interruptions

            of people in human-computer interaction. Human-Computer Interaction, 17, 63-139.

Mischel, W.,  & Masters J. C. (1966).  Effects of probability of reward attainment on responses

            to frustration.  Journal of Personality and Social Psychology, 3,  390-396. 

North, D. W. (1968). “A tutorial introduction to decision theory.” IEEE Trans. Systems Science

and Cybernetics, (4). Reprinted in Shafer & Pearl.

Nygren, T. E., (1998).  Reacting to perceived high-and low-risk win-lose opportunities in a risky

   decision-making task:  Is it framing affect or both? Motivation and Emotion,  22, 73-98.

Schlottman, A., & Tring, J.  (2005).  How children reason about gains and losses: Framing

            effects in judgment and choice.  Swiss Journal of Psychology, 64,  153-171.

Seshadri, S., & Shapira, Z. (2001). Managerial allocation of time and effort: The effects of

            interruptions. Management Science, 47, 647-662.

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.

            Science, 185, 1124-1131.

Tversky, A., & Kahneman, D. (1988). Rational choice and the framing of decisions.  In Bell, D.

            E., Raffia, H., & tversky, A. (Eds) Decision making: Descriptive, normative, and

            prescriptive interactions. New York, NY, US: Cambridge University Press.

Williams, S., & Voon, Y. W. W. (1999).  The effects of mood on managerial risk perceptions:

Exploring affect and the dimensions of risk.  The Journal of Social Psychology,  139,  268-287. 

Xia, L., & Sudharshan, D. (2002). Effects of interruptions on consumer online decision

 processes. Journal of Consumer Psychology, 12, 265-280.

Xie, X-F., & Wang, X. T. (2003). Risk perception and risky choice: Situational, informational and dispositional effects. Asian Journal of Social Psychology, 6, 117-132.

Table 1     Degrees of Freedom, F Values, and Significance Values

                                                                _df__                          F                    sig

Main Effect for Framing___________________1,106___________13.207______<.001*

Main Effect for Frustration_________________1,106___________0.361_________.055

Main Effect for Interruption________________1, 106___________4.633_______0.034*

Framing x Frustration Interraction___________1,106____________4.633_______0.034*

Framing x Interruption Interaction___________1,106____________0.361_________.055

Frustration x Interruption Interaction_________1,106_____________2.39________0.125

Framing x Frustration x Interruption Interaction_1,106____________2.809_______0.097

                                                                             *Values are significant at .05 level

            Table 2_____Mean and SD for Not Interrupted and  Interrupted and t values

                 Not Interrupted (Mean/SD)                 Interrupted (Mean/SD)             t value

Anxiety___________2.19/1.135___________________2.52/1.208______t(109)=1.432, p=.155

Frustration________ 1.94/1.111___________________2.58/1.467______t(109)=2.515, p=.013

Attention__________3.28/1.246___________________3.05/1.133______t(109)=1.012, p=.314

Wording Impact_____1.8/.749_____________________1.81/.618______ t(107)=.039, p=.969

Time Impact________2.39/1.406___________________3.19/1.402_____ t(108)=2.935, p=.004

Table 3__Mean and SD for Easy and Hard Math Tests and t values

            Easy (Mean/SD)                        Hard (Mean/SD)                  t value

Anxiety_____2.08/1.062________________2.73/1.234________t(109)=2.946, p=.004

Frustration___1.72/.865_________________3/1.51___________t(109)=5.593, p=.000

Attention____3.13/1.2__________________3.16/1.173________t(109)=.104, p=.917

Wording Impact__1.83/.673_____________ 1.78/.679________  t(107)=.389, p=.689