Walker, I. (2006). Making Immediate Decisions Between Mutually Exclusive Options. PHILICA.COM Article number 2.
Making Immediate Decisions Between Mutually Exclusive Options

Ian Walkerconfirmed userThis person has donated to Philica (Department of Psychology, University of Bath)

Published in psycho.philica.com

Abstract
An experiment looked at how people make decisions between two mutually exclusive options (here, between the responses ‘yes’ and ‘no’). The logic of this decision is different to those normally studied in naturalistic decision-making research, as to reject one option is effectively to accept the other. The aim was therefore to see whether such a decision is still made by considering options in series, as in other decisions. The procedure biased people’s expectations about the relative proportions of ‘yes’ and ‘no’ responses they would have to make. The results showed that this type of decision is indeed made by considering the options in series. The finding that even this most simple of decisions is made serially suggests that human decision making in bounded rationality situations is intrinsically serial. However, a question raised is whether the characteristics of mutually exclusive decisions might alter the nature of the serial decision process itself.

Introduction

Simon (1957) introduced the bounded rationality approach to decision-making, which describes decision processes in realistic situations of limited resources, e.g., limited time, information, or processing capacity. The basis of decision-making as envisaged by Simon is that options are evaluated in series and each is compared to some criterion of goodness. This criterion will vary depending upon the task in hand and can even vary from moment to moment (Sivak, 2002). The first option to be considered that satisfies the criterion will be the one selected.

This serial aspect of Simon's account has proved influential, and other models of decision making have tended to describe a similar mechanism (Hammond, 1988; Klein, 1993; Lipshitz, 1993; Rasmussen, 1983). However, given this view that decision-making involves a serial appraisal of options, a question we might ask is whether such an approach is applicable to all decisions or whether we might have to appeal to different, non-serial mechanisms for certain types of decision. In particular, what are the cognitive processes involved when people must decide between two mutually exclusive options (MEOs), e.g., ‘yes' and ‘no'?

Although a choice between two MEOs is perhaps the simplest form of decision that can be made, it appears a priori to be different to other decisions. The sort of decision usually considered in naturalistic decision-making research involves a choice between several options with no mutual exclusivity. For example, imagine the situation in which somebody is required to choose a course of action in almost any field (Klein, 1993). Here, to reject the first option does not necessarily do anything to make the acceptance of any of the others more likely. The decision made in a yes/no choice, on the other hand, does have this property, as there is only one degree of freedom and to reject one option is effectively to accept the other. It seems reasonable to ask, then, whether we still make such decisions using serial mechanisms, given that serial consideration is not necessary, and indeed might not be the best way to make such a decision. This question was investigated by giving people a simple yes/no decision and measuring their response times in a way that allowed us to see whether they were using serial decision processes or not.

This question of how people choose between two MEOs was first raised in a recent study by the author (Walker & Brosnan, submitted), where drivers saw photographs of a cyclist signalling at a T-junction and had to make a yes/no judgement about whether the cyclist was about to turn the corner or not. That experiment did not allow us to decide whether the decision was made serially (it was not intended to, incidentally). The participants there, on average, responded ‘yes' about as quickly as they responded ‘no'. On first examination, these results might therefore seem to favour an account in which the two options are considered in parallel, but this conclusion is not safe. The participants responded ‘yes' and ‘no' equally quickly, but they also responded ‘yes' and ‘no' equally often. Therefore, the data are also consistent with a serial account, provided we make the reasonable assumption that people were considering ‘yes' then ‘no' on half the trials and ‘no' then ‘yes' on the other half. The two would therefore average out, giving equal times for ‘yes' and ‘no' responses.

The present experiment attempted to answer the question of how people choose between MEOs whilst addressing this issue. It asked participants to make a similar type of judgement, but biased their expectations. Specifically, participants were asked to make a simple yes/no decision but were informed at the start of the experiment what the overall proportions of ‘yes' and ‘no' responses would be during the procedure. One-third of the participants needed to respond ‘yes' on 30% of trials, another third needed to respond ‘yes' on 70% of trials, and the final third (the baseline group) needed to respond ‘yes' 50% of the time.

If this type of decision is not made serially, then biasing the procedure in this way will have no effect on people's reaction times for ‘yes' and ‘no' responses - there will be the same difference between the two (most likely, a very small difference) for all three groups. If the serial account is correct, on the other hand, then this will not be the case. If participants are making their decisions by considering the two options in turn, it seems reasonable to assume that people who know they will mostly have to make ‘yes' responses will consider this option first most (or all) of the time. Therefore, when a ‘yes' response is required they will tend to make it quickly; when a ‘no' response is required they will tend to make it more slowly because it will be the second option considered. Similarly, participants who are mostly making ‘no' responses should be slower to make ‘yes' responses than ‘no' responses. The specific hypothesis under test, then, is that the 30%-yes group will be faster to make ‘no' responses than ‘yes' responses whereas the 70%-yes group will be faster to make ‘yes' responses than ‘no' responses; the 50%-yes group will fall in between. If this pattern is not found and all the groups respond ‘yes' and ‘no' in the same way, this will show that a decision between two MEOs is instead made by considering them both simultaneously.

Method

Participants
42 psychology students from a UK university volunteered to take part in this experiment. There were 34 women and 8 men (which is typical for UK psychology classes) and ages ranged from 19 to 26 (mean 19.74). Problems with data files meant that the data from 3 participants could not be used.

Procedure
The experiment was controlled by the E-Prime package (Psychology Software Tools, 2003). The scripts were made available to the participants so that they could complete the procedure at the end of a lecture in a university computer laboratory. Participants worked independently and the experiment lasted around three minutes.

The participants were split into three groups by the experimenter and each used a slightly different version of the script. There was a 30% Group, a 50% Group, and a 70% group, named to reflect the proportion of ‘yes' responses that the participants would be required to produce during the procedure.

The experiment began with written instructions on the computer's screen, which described the procedure and told the participants the proportion of ‘yes' and ‘no' responses they should expect to make (30%, 50%, or 70%). Each trial consisted of a fixation point presented on the computer's screen for 1 s. This was then replaced by a large letter X, which was either red or blue. Whether the X was red or blue on a particular trial was random for the 50% group and pseudo-random for the other groups (i.e., there was no fixed pattern from trial to trial, but across the whole procedure the proportion of red X's was 30% or 70%). Participants answered as quickly as possible the question ‘Is the X red?', and pressed ‘y' or ‘n' on the computer's keyboard to indicate their decision. After the response, there was an interval of 1.5 s before the next trial began. There were 50 trials in total.

Results

Mean reaction times for ‘yes' and ‘no' responses, as well as the number of errors made, were collected from each participant. The mean reaction times to provide ‘yes' and ‘no' responses for each group is shown in Figure 1. From this it can be seen that overall, participants made their decisions fairly rapidly. It can also be seen that the predicted pattern of results was found: participants in the 30%-yes group were faster to make ‘no' responses than ‘yes' responses, showing that they were tending to consider ‘no' responses first. Participants in the 70%-yes group were faster to make ‘yes' responses than ‘no' responses, showing that they were considering ‘yes' responses first. The 50%-yes group falls somewhere between these two extremes, as we would expect, although the particular pattern of results obtained for this group is interesting as it suggests that peoples' default behaviour in this task is to consider ‘yes' before ‘no' to some extent (this would also explain why the difference for the 30% group is smaller than the difference for the 70% group). A two-way split-plot analysis of variance showed that there was a significant main effect on reaction times from Answer (F(1,36) = 6.52, MSE = 407.54, p < .02, ηp2 = 0.15) but not from Group (F(2,36) = 0.25, MSE = 7296.36, ns, ηp2 = 0.01). However, as expected from Figure 1, the Answer x Group interaction was significant (F(2,36) = 5.02, MSE = 407.54, p = 0.01, ηp2 = 0.22). This effect size is small, but large enough to confirm that this is a real, consistent effect.

Figure 1 — Mean times to provide ‘yes’ and ‘no’ responses by group


The mean number of errors (and standard deviations) for each group were as follows: 30% group = 1.00 (1.04), 50% group = 1.17 (1.59), 70% group = 1.08 (1.04). Clearly, then, participants made very few errors in this procedure. One-way independent-measures analysis of variance confirmed that the three groups did not differ significantly in their error scores (F(2,36) = 0.09, MSE = 1.52, ns, ηp2 = 0.00).

Finally, an analysis was carried out to check that these effects were caused by the global biases given to the participants and not by local information available to the participants during the procedure. That is, the 30% group were mostly experiencing trials in which they had to respond ‘no' and the 70% group were mostly experiencing trials in which they had to respond ‘yes'. Therefore, these people particularly would have often experienced sequences of trials all requiring the same response. Although it would not change our basic finding that the MEO decision process is serial, an alternative explanation for the results in Figure 1 might be that because the 30% group would often have a run of trials in which they had to respond ‘no' each time, this (i.e., their experience of the last few trials) might be what was making them consider ‘no' first, rather than the global bias introduced by the experimenter. A similar argument can be made for the 70% group. Therefore an analysis was carried out on the runs in the data. A run was defined as a sequence of 4 or more trials that required the same response, and importantly a participant responding one trial at a time would have no way of knowing when a run would end. There were 117 runs in total and these were analysed in two ways. First, for each run, the difference in response time between the first and last trial was measured. If it was simply the experience of the last few trials that was causing the seriality seen in Figure 1 then we should expect trials at the end of a run to have noticeably different response times than trials at the beginning. Second, the difference in response time between the last trial in each run and the first different trial after the run was measured. If it was local information (i.e., the experience of the last few trials) that was biasing participants' decisions then there should be a notable difference in response time between these (e.g., if, after several ‘no' responses the participant began to expect a ‘yes', they would be rather faster on the first different trial; alternatively if, after several ‘no' responses the participant was expecting another ‘no', they would be rather slower on the first different trial).

Overall, the data suggest that it was not local information (i.e., the experience of the last few trials) that was causing participants to consider one option before the other. The mean difference between the first and last trial in a run was 19.71 ms (SD = 104.02, range = -486 to 364) and the mean difference between the last trial of a run and the first trial after the run was only -9.99 ms (SD = 135.43, range = -356 to 561). If participants were being biased to consider one option ahead of another based on their experiences of the past few trials, rather than on the global information about the distribution of responses, we should expect these differences to be considerably more extreme. The standard deviations and ranges show that there was great variation in these measures, confirming that there is no consistent effect of participants' recent experiences on their tendency to consider one response ahead of the other in any given trial. This must then have been caused by the bias information given at the start of the experiment.

Discussion

This experiment looked at how people make an immediate decision between two mutually exclusive options (MEOs). Specifically, given that such decisions are extremely simple and have certain logical differences from the decisions usually considered by researchers in decision-making, the aim was to see whether they are still made serially, like other naturalistic decisions.

The data presented here suggest that when people choose between two MEOs they do consider one option before the other. In other words, when choosing between two options, where to reject one is effectively to accept the other, people still use serial mechanisms. This finding is unlikely to have been an experimental artefact caused by the biasing manipulation, as the people in the control group, who made an unbiased decision, also seemed to exhibit seriality in their decision processes.

Given the simplicity of the decision in this experiment, the results suggest that serial mechanisms might be the only ones available to people making decisions in bounded rationality situations. If people do make all such decisions serially, a question that arises is whether the nature of the serial decision process itself can ever change. Specifically, given the properties of a mutually exclusive decision, in which to reject one option is effectively to accept the other, we can ask whether the second option is considered fully, or whether rejecting the first automatically triggers the acceptance of the second without it being fully considered. A model in which the second option is considered fully if the first is rejected (as in standard models of naturalistic decisions) might be termed the strong serial account; a model in which the first MEO is fully considered but the second automatically accepted if the first is rejected might be termed the weak serial account. The strong serial account allows for the possibility that people might find that both options do not meet their decision criterion, in which case they would presumably choose the least unsatisfactory option or perhaps refuse to make a decision. The strong serial account therefore allows for the possibility that the decision maker can introduce further de facto options into a choice between two MEOs. The data from the present experiment do not allow us definitely to choose between the strong and weak accounts, and further work will be needed to see whether people are sensitive to mutual exclusivity in such a way that they shift to a weak serial mechanism when the decision has this property. Although having said this, the relatively large difference between ‘yes' and ‘no' response times for the 70% group (Figure 1) perhaps suggests that the ‘no' response was not automatically being accepted upon ‘yes' being rejected. Another task that arises from this study is to see how these results change in situations where there are three or more MEOs.

References

Hammond, K.R. (1988). Judgement and decision making in dynamic tasks. Information and Decision Technologies, 14, 3-14.

Klein, G.A. (1993). A recognition-primed decision (RPD) model of rapid decision making. In G.A. Klein, J. Orasanu, R. Calderwood, & C.E. Zsambok (Eds), Decision Making in Action: Models and Methods. (pp. 138-147). Norwood, NJ: Ablex.

Lipshitz, R. (1993). Converging themes in the study of decision making in realistic settings. In G.A. Klein, J. Orasanu, R. Calderwood, & C.E. Zsambok (Eds), Decision Making in Action: Models and Methods. (pp. 103-137). Norwood, NJ: Ablex.

Psychology Software Tools, Inc. (2003). E-Prime [computer software].

Rasmussen, J. (1983). Skill, rules, and knowledge: Signals, signs and symbols, and other distinctions in human performance models. IEEE Transactions on Systems, Man and Cybernetics, 13, 234-243.

Simon, H.A. (1957). Models of man: Social and rational. New York: Wiley.

Sivak, M. (2002). How common sense fails us on the road: contribution of bounded rationality to the annual worldwide toll of one million traffic fatalities. Transportation Research Part F, 5, 259-269.

Walker, I. & Brosnan, M. (in prep). Drivers' interpretations of a cyclist's gaze and arm signals in a simple judgement task.

Information about this Article
Peer-review ratings as of 11:47:15 on 23rd Oct 2017 (from 3 reviews, where a score of 100 is average):
Originality = 148.17, importance = 116.73, overall quality = 132.95

Published on Sunday 23rd April, 2006 at 16:50:12.

Creative Commons License
This work is licensed under a Creative Commons Attribution 2.5 License.
The full citation for this Article is:
Walker, I. (2006). Making Immediate Decisions Between Mutually Exclusive Options. PHILICA.COM Article number 2.

Peer review added 3rd May, 2006 at 08:38:48

The article is pretty interesting. Manipulating the expectation of the participant before the decision task with a simple instruction seems to have significantly influenced their Yes/No decision.

However, I would say that the conditions here actually indicate that the decision making process is not, as the authors describe, a serial process. If it is, then the decision to reject one option rather than accept the other (a caution decison maybe?) can be influenced.

What if a third option is introduced into the process? Here we have a yes/no decision. What of a third option where participants can indicate their certaintly, perhaps a measure of risk in the process?

However, that is not to detract from the result, which seems an interesting step in exploring this issue. Potentially useful.

Peer review added 6th October, 2006 at 04:12:27

This article reminds me of my childhood days, when I would choose the last option. If someone said “Yes or No?”, I would choose “No”, while if someone said “No or Yes”, I would say “Yes.”.

Perhaps that was because I only remembered the most recent option.

Similarly, I feel that the order of choices presented, may make a small difference in what people choose. Some people tend to like the first choice, no matter what the circumstances, or some people tend to like the latter choice.




Website copyright © 2006-07 Philica; authors retain the rights to their work under this Creative Commons License and reviews are copyleft under the GNU free documentation license.
Using this site indicates acceptance of our Terms and Conditions.

This page was generated in 0.0351 seconds.