The purpose of the Cengage CogLab Risky Decisions experiment (RDE) had been to demonstrate framing effects which lead to violation of rational criterions of consistency and coherence (Kahneman & Tversky, 1982; Tversky & Kahneman, 1981). Before unpacking framing effects, a decision frame is effectively a mental set of “acts, outcomes, and contingencies associated with a particular choice” (Tversky & Kahneman, 1981, p. 453). Frames may be selected (in decisions) based on “formulation of the problem and partly by norms, habits, and personal characteristics of the decision-maker” (p. 453). In effect, schemas, procedural memory, and traits (possibly inclusive of moods, and emotion) influence which frame is selected out of a set of frames.

It is rationalized that changing frames should not change while changing perspective. Literature provided an example of veridical perception where changes of the vantage point while looking at two mountains should not change the perceived relative height (Tversky & Kahneman, 1981). Framing effects had been evidenced to influence individuals to (a) avoid risk when there are potentials for gains, (b) prefer more certain gains vs. uncertain (i.e., risky) losses, (c) seek risk when there are potentials for losses, and (d) prefer uncertain (i.e., risky) loss vs. certain loss. Setting aside clear issues of sensory and short-term memory decay, perspective, and occlusion where working memory may not be able to chunk the frame set efficiently (if at all), absolute heights of said mountains may appear differently due to perspective from vantage points, and objects in the foreground may occlude object bases in the background. It is easy to understand, at face value, how a reversal of frames could occur.

Hypotheses

Though the RDE did not offer a hypothesis, deducing hypotheses from CogLab’s own provided experiment background and literature review reveals that it had been hypothesized that experimenters could manipulate participants such that individuals will reverse frames (i.e., acts, outcomes and contingencies) depending on a perspective offered. Unfortunately, this “hypothesis” is a set of hypotheses, but thankfully already mentioned above in anticipated framing effects: individuals will (a) avoid risk when there are potentials for gains, (b) prefer more certain gains vs. uncertain (i.e., risky) losses, (c) seek risk when there are potentials for losses, and (d) prefer uncertain (i.e., risky) loss vs. certain loss. If reversal of frames had not been evidenced via these hypotheses, then the aggregate hypothesis of framing reversal would be rejected. 

It is worth noting that the RDE experiment is much more complex in hypothesis under the hood and requires additional participant skills in mathematics which would far outstrip the length of this paper (even the Scientific American article is approximately 6 pages of extremely small print and does not cover the full math background required). The reason for this is that certain combinations proportions of risky bets, and guaranteed gains may result in similar probabilities which are modified by perception into convex/concave value functions (Tversky & Kahneman, 1982). The original premise of investigations into risky decisions goes deeper, as the original hypothesis investigated the idea of human rationality but is obfuscated by concomitant work to determine optimal risk which requires special calculations (see discussion).

Independent/Dependent Variables

The RDE consisted of multiple independent variables in a 2 × 2 factorial design containing 2 conditions of guaranteed “small gain” or “large gain”, each with two levels of a risky bet divided into “less risky” and “more risky”. The dependent variable had been the proportion of times a participant chose either to gamble on a less- or more- risky bet vs. take the guaranteed small/large gain.

Method/Procedure

The 2 × 2 factorial design presented a series of trials where participants were presented on a computer screen a “wheel of fortune” type game which presented (a) a wallet (i.e., accumulator) containing the total dollar amount the participant has, (b) two conditions of guaranteed gains bottom right with random values assigned to each condition, (c) two levels of risky bets bottom left consisting of a random amount to win or lose along with respective probabilities assigned to each level, (d) a “spinnable” wheel representing a pie representing the respective volumes of each level’s probability, and (e) a pointer indicating the place on the wheel to measure for evaluating the result of an executed risky bet. If the participant chose the guaranteed gain (or technically a negative gain), the value is added to the wallet. If the participant chose the risky bet, the wheel is spun, slows down (this part of the experiment is not well described, and had been observed to be variable, technically adding another set of levels, but for this paper had been assumed to follow exactly aforementioned probability), and when the wheel stops, if the pointer is on green, the amount won is added, otherwise if red, the amount lost is added. The initial starting condition sets the wallet at $1000, and each participant executes 32 trials.

Results

Group data (N = 24) partially confirms the hypothesis that framing effects reversed decision frames. To borrow from Tversky and Kahneman (1981)’s hypotheses with a mapping to RDE results (in italics), (a) participants avoided risk when there had been potentials for gains; participants gambled the least in the face of larger gains, but significantly more in the less-risk/small-gain scenario, (b) participants preferred more certain gains vs. uncertain (i.e., risky) losses; participants gambled less in the face of more risk, (c) participants sought risk when there had been potentials for losses; participants gambled more when there were small/large losses, but significantly less than the less risk/small-gain scenario, and gambled least in the more risk/small-loss scenario, and (d) participants preferred uncertain (i.e., risky) loss vs. certain loss; participants gambled most in the large-loss/more-risk condition/level in the more risky condition, and gambled in the second most in the large-loss/less-risk condition/level in the less risky condition.

In analytical conclusion, the results partially disconfirmed the hypothesis, which implies a partially confirmed hypothesis. However, reversal is clearly seen in the small- and large- gain/less-risk conditions/level vs. the small- and large- gain/more-risk conditions/level. In the former, participants gambled more in the small-gain/less risk condition/level vs. more in the large-gain/more-risk condition/level. So, there was some reversal, but there had been some inconsistencies regarding the hypotheses. These results may be a bit inconclusive, but more likely due to experimental design. Due to the experiment’s design, the framings of equal probabilities had not been evaluated, therefore, limiting evaluation of the core hypothesis. Unfortunately, a lay description would be thus: this experiment neither confirmed nor denied the absent stated hypothesis associated with the RDE, yet partially confirmed partially the hypothesis of framing reversal in a deduced hypothesis through satisfaction of a larger portion of the hypotheses—a risky bet indeed.

Discussion

The original paper explaining the underlying phenomena of “risky decisions” asserts that human decisions are not necessarily influenced by consideration of pure measures of probability, and that mathematical functions of decisions, subject to individual differences, may be modeled (Kahneman & Tversky, 1982). In discussion, Kahneman and Tversky draw on Herbert Simon’s (1957) concept of bounded rationality which considers the “replace[ment of] the perfect rationality assumptions of homo economicus with a conception of rationality tailored to cognitively limited agents” (Wheeler, 2018; edited for readability). While somewhat insulting, cognitive limitation is subject to cognitive load. Essentially, as the calculation of probabilities of a system requires skill in not only theory but mathematical calculation, humans may be effectively limited in cognitive capacities related to these calculations, which is subject of an entire nature/nurture debate. In addition, salient stimuli in an environment may interrupt, distract, and/or stress a participant during calculation. There is also a whole matter of stereotype threat to consider. As heuristics are drawn more frequently under high cognitive load (e.g., Brunyé et al, 2017), it is arguable that “cognitively limited” is more related to limits of cognitive load. It is also uncertain if researchers considered limitations of viewpoint invariance where objects are occluded, especially if it is the first time seeing said object—limiting calculations of probability due to time, resource, or effects of psychologically known effects seems a bit of a confound.

With respect to the experiment’s design, the partial confirmation of the hypothesis may contribute to the field by calling for improvements to RDE experimental design by more closely operationalizing a hypothesis or hypotheses with respect to theoretical constructs related to framing effects. For example, with respect to literature, it is written, “framing effects arise when the same objective alternatives are evaluated in relation to different points of reference” (Kahneman & Tversky, 1982, p. 166). Therefore, it seems more effective to measure gambling behaviors with respect to the “same objective alternatives” as calculated. However, the problem still does not solve the issue with the entire construct. That is effectively this: what is the realistic expectation of a biological neural network not equipped with a mathematical computational machine which calculates minimization of risk and maximization of returns. If humans had this, “all humans would win”, and be kicked out the market, until new rules loaded humans such that again, working memory is overloaded, leading to heuristics, burnout, and just plain guessing—enter the Vegas casino.

As is also said, “framing effects in consumer behavior may be particularly pronounced in situations that have a single dimension of cost (usually money) and several dimensions of benefit” (Kahneman & Tversky, 1982. p. 168). Unfortunately, the RDE did not evaluate this hypothesis either. Risky decision literature is loaded with hypotheses, for example, another one, “… the threat of a loss has a greater impact on a decision than the possibility of an equivalent gain” (Kahneman & Tversky, 1982, p. 160)—again that word “equivalent” is part of an operationalization of experimentation. Another hypothesis, “the regret associated with a loss that was incurred by an action tends to be more intense than the regret associated with inaction or a missed opportunity” (p. 160), while certainly interesting, this is also untested. So, the jury’s out which hypothesis was being tested in the RDE. It was almost as if the experiment was chasing after a hypothesis that fit its evidence. The experiment may simply have been rushed, or its underlying hypothesis unknown by this author.

That all said, risky decision experimentation and literature is powerful, and is rooted deep in history because it is based on Daniele Bernoulli’s monumental yet invisible work (Stearns, 2000). Bernoulli’s (1738) paper, while ignored by mathematicians and physicists had contributed tremendously to theories of risk management in economics and evolution (Stearns, 2000). The method Bernoulli used to minimize risk had been to measure risk with a geometric mean and spread it across “a set of independent events… bet-hedging” (Stearns, 2000). Bernoulli practically invented a method to manage risk, and it is core to many theories. This work is combined with Kahneman & Tversky’s literature, and merges economics with psychology. To that end, as this work is related to bet hedging, it is most likely being abused, and this experiment partially demonstrates how.

The hypotheses partially confirmed in experimentation implies an ethical dilemma that is profound. If psychologists learned that economists could manipulate public behavior to not choose the most optimal method to minimize risk and to cheat the public out of that which is exchanged for improving health and well-being (i.e., notes denominating debt accrued [i.e., currency]), then that would be a violation of ethics. If the public is simply being intentionally cognitively overloaded by more complex problems than individuals in the public can chunk in a normalindividual’s working memory while simultaneously taking advantage of psychological phenomena such as embarrassment avoidance in high public self-conscious individuals shopping in public areas or participating in game shows in public (see Fenigstein et al., 1975, 1984; von Gemmingen et al., 2003) etc., then who is policing basic and applied psychological research in the economists’ domain? Are imperialist ethnocentric ideas based in genetic/moral superiority still excusing these behaviors?

Given the increases in income and wealth inequalities, it perhaps is no surprise that the scientific field, and its associations have fell asleep at the wheel, while theories have been applied for those with more equipment to calculate more and more complex phenomena while humans less privileged are limited by not only working memory, but training in more and more effective chunking through affordances of time, space, money, health, well-being, and a variety of other factors in development such that they can have a fair shot acquiring the same tools. Perhaps the real implication is thus: scientists may be too embarrassed and too beholden to the very market abused, that they themselves are paralyzed by strain adaptations functionally fixated on the scientific method, that they had forgotten its whole point. Are scientists too afraid to police their own applications. Perhaps it’s time for the Mertonian stressed and strained “cognitively limited agents” to wake up. It’s right here!

References

Bernoulli, D. (1738). Specimen theoriae novae de mensura sortis. Commentarii academiae scientiarum imperialis Petropolitanae, 5, 175-192. https://archive.org/details/SpecimenTheoriaeNovaeDeMensuraSortis/

Brunyé, T. T., Martis, S. B., & Taylor, H. A. (2017). Cognitive load during route selection increases reliance on spatial heuristics. Quarterly Journal of Experimental Psychology, 71(5), 1–38. http://doi.org/10.1080/17470218.2017.1310268

Fenigstein, A. (1984). Self-consciousness and the overperception of self as a target. Journal of Personality and Social Psychology, 47(4), 860. https://doi.org/10.1037/0022-3514.47.4.860

Fenigstein, A., Scheier, M. F., & Buss, A. H. (1975). Public and private self-consciousness: Assessment and theory. Journal of Consulting and Clinical Psychology, 43(4), 522. https://doi.org/10.1037/h0076760

Khaneman, D., & Tversky, A. (1982). The psychology of preferences. Scientific American, 246(1), 160–173. https://doi.org/10.1038/scientificamerican0182-160

Simon, H. A. (1957). Models of man. John Wiley.

Stearns, S. C. (2000). Daniel Bernoulli (1738): Evolution and economics under risk. Journal of Biosciences, 25, 221–228.

Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458. https://doi.org/10.1126/science.7455683

von Gemmingen, M. J., Sullivan, B. F., & Pomerantz, A. M. (2003). Investigating the relationships between boredom proneness, paranoia, and self-consciousness. Personality and Individual Differences, 34(6), 907–919. https://doi.org/10.1016/S0191-8869(01)00219-7

Wheeler, G. (2018). Bounded rationality. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/bounded-rationality/