Decision inertia is the tendency to repeat previous choices independently of the outcome, which can give rise to perseveration in suboptimal choices. We investigate this tendency in probability-updating tasks. Study 1 shows that, whenever decision inertia conflicts with normatively optimal behavior (Bayesian updating), error rates are larger and decisions are slower. This is consistent with a dual-process view of decision inertia as an automatic process conflicting with a more rational, controlled one. We find evidence of decision inertia in both required and autonomous decisions, but the effect of inertia is more clear in the latter. Study 2 considers more complex decision situations where further conflict arises due to reinforcement processes. We find the same effects of decision inertia when reinforcement is aligned with Bayesian updating, but if the two latter processes conflict, the effects are limited to autonomous choices. Additionally, both studies show that the tendency to rely on decision inertia is positively associated with preference for consistency.
1. INTRODUCTION
As described in Newtonian physics, the term “inertia” refers to the fact that, in the absence of external resistance, a moving object will keep moving in the same direction. This word has also been used across multiple fields as a metaphor to describe related characteristics of human behavior. For example, in management and organization science, the expression “cognitive inertia” describes the phenomenon that managers might fail to reevaluate a situation even in the face of change (Huff et al., 1992; Reger and Palmer, 1996; Hodgkinson, 1997; Tripsas and Gavetti, 2000). In medical studies, “therapeutic inertia” or “clinical inertia” describe the failure of health care providers to intensify therapy when treatment goals are unattained (Phillips et al., 2001; Okonofua et al., 2006). In sociology, “social inertia” depicts the resistance to change or the (excess) stability of relationships in societies or social groups (Bourdieu, 1985). In psychology, the “inertia effect” describes individuals’ reluctance to reduce their confidence in a decision following disconfirming information (Pitz and Reinhold, 1968). The concept of “psychological inertia” has been proposed to describe the tendency to maintain the status-quo (Gal, 2006). Suri et al. (2013) speak of “patient inertia” to describe the phenomenon that many patients stick to inferior options or fail to initiate treatment even after the diagnosis of a medical problem.
Summing up, the concept of inertia has been used to describe many different phenomena related to a resistance to change. The existence of these phenomena has been linked to status-quo bias (Samuelson and Zeckhauser, 1988; Ritov and Baron, 1992), described as the tendency to maintain the defaults either by repeating a decision or avoiding action. So far, however, our understanding of the processes underlying inertia in decision making is rather limited. In the present study, we aim to contribute to this understanding by focusing on a particular facet of inertia, which we term “decision inertia:” the tendency to repeat a previous choice, regardless of its outcome, in a subsequent decision. We investigate whether this tendency significantly influences active decision making and explore the psychological processes behind it using a belief-updating task.
The phenomenon we explore here is consistent with previous evidence from the decision-making literature. For instance,Pitz and Geller (1970) observed a tendency to repeat previous decisions even following disconfirming information. In a study on reinforcement in belief-updating tasks, which was not focused on inertia, Charness and Levin (2005) nevertheless observed a “taste for consistency,” corresponding to the phenomenon that people were prone to repeat their choices, no matter whether these choices led to success or failure. In a study on perceptual decision making, Akaishi et al. (2014) showed that choices tend to be repeated on subsequent trials, even on the basis of little sensory evidence. Erev and Haruvy (in press) review studies on decision making from experience where, for instance, participants repeatedly choose between a risky prospect and a safe option, and receive immediate feedback (e.g., Nevo and Erev, 2012). Erev and Haruvy (in press) conclude that there exists a strong tendency to simply repeat the most recent decision, which is even stronger than the tendency to react optimally to the most recent outcome. Furthermore, Zhang et al. (2014) showed that the tendency to repeat previous decisions exists even for unethical behavior. There might also be a relation to the extensive literature on choice-induced preference change, which shows that earlier decisions alter preferences, and hence might result in repeated choices (see also Ariely and Norton, 2008; Sharot et al., 2010; Alós-Ferrer et al., 2012).
The influence of previous decisions on subsequent choices has also been investigated in reinforcement learning research. For instance, Lau and Glimcher (2005) studied trial-by-trial behavior of monkeys in a matching task in which the reward structure favored alternating between two choice options. They observed that choosing a particular alternative decreased the probability of choosing that alternative again on the next trial, but increased the likelihood of choosing it again some time in the future, regardless of reward history. Studies in which participants worked on probabilistic “bandit tasks” that favored sticking with successful options showed that participants were prone to repeat their choices, independently of any effects due to previous rewards (e.g., Schönberg et al., 2007, Supplemental Results). Accordingly, reinforcement learning models now account for the influence of past choices on subsequent ones by including a model parameter of “perseveration,” capturing the tendency to repeat or avoid recently chosen actions (e.g., Schönberg et al., 2007; Gershman et al., 2009; Wimmer et al., 2012; for an introduction to model- based reinforcement learning, see Daw, 2012). The inclusion of such a parameter leads to more accurate predictions in contrast to models that merely incorporate the effect of past reinforcers (see Lau and Glimcher, 2005).
To understand decision inertia, we consider a multiple- process framework (Evans, 2008; Sanfey and Chang, 2008; Weber and Johnson, 2009; Alós-Ferrer and Strack, 2014), that is, we consider individual decisions as the result of the interaction of multiple decision processes. Specifically, we follow the assumptions of parallel-competitive structured process theories, which propose that multiple processes affect behavior simultaneously, resulting in conflict or alignment among these processes (e.g., Epstein, 1994; Sloman, 1996; Strack and Deutsch, 2004). Whenever several decision processes are in conflict (i.e., deliver different responses), cognitive resources should be taxed, resulting in longer response times and higher error rates. These predictions were confirmed in a response-times study by Achtziger and Alós-Ferrer (2014), which showed that more errors arise and responses are slower when Bayesian updating (i.e., normatively optimal behavior) is opposed to reinforcement learning of the form “win-stay, lose-shift.” We relied on a variant of the experimental paradigms employed in Achtziger and Alós- Ferrer (2014), Achtziger et al. (2015), and Charness and Levin (2005) but focused on the conflict with decision inertia, viewed as a further decision process. We measured error rates and response times to investigate the role of decision inertia in a belief- updating task. Specifically, we hypothesized that decision inertia is a further process potentially conflicting with optimal behavior and affecting decision outcomes and decision times. Accordingly, our main hypotheses were that more errors and slower choices would be made in cases of conflict between decision inertia and Bayesian updating.
To further explore decision inertia, we considered possible individual correlates of this decision process. We hypothesized that decision inertia would be associated with preference for consistency (PFC), which is a desire to be and look consistent within words, beliefs, attitudes, and deeds, as measured by the scale with the same name (Cialdini et al., 1995). Cialdini (2008)argues that because of the tendency to be consistent, individuals fall into the habit of being automatically consistent with previous decisions. Once decision makers make up their minds about a given issue, consistency allows them to not think through that issue again, but leads them to fail to update their beliefs in the face of new information when confronting new but similar decision situations. Furthermore, Pitz (1969) observed that inertia in the revision of opinions is the result of a psychological commitment to initial judgments. Thus, we hypothesized that preference for consistency might be one of the possible mechanisms driving decision inertia, in which case an individual’s behavioral tendency to rely on decision inertia should be positively associated with their preference for consistency (PFC).
Our last hypothesis concerns the kind of decisions leading to decision inertia. If this phenomenon arises from a tendency to be consistent with previous decisions, and hence economize decision costs, the effect should be stronger following autonomous decisions (free choices) than required ones (forced choices). The same prediction also arises from a different perspective. In general, human decision makers prefer choice options that they freely chose over options with equal value that they did not choose, as exemplified by the literature on choice-induced preference change (e.g., Lieberman et al., 2001; Sharot et al., 2009, 2010; Alós-Ferrer et al., 2012). Relying on behavioral and genotype data, Cockburn et al. (2014) recently investigated the underlying mechanism of this preference. In a probabilistic learning task, their participants demonstrated a bias to repeat freely chosen decisions, which was limited to rewarded vs. non-rewarded decisions. Interindividual differences in the magnitude of this choice bias were predicted by differences in a gene that has been linked to reward learning and striatal plasticity. Cockburn et al. (2014) interpret these findings as evidence that free choices selectively amplify dopaminergic reinforcement learning signals, based on the workings of a feedback loop between the basal ganglia and the midbrain dopamine system. Given such an amplification of the value of freely chosen options, it again follows that decision inertia should be more pronounced after autonomous decisions compared to forced ones in our study. We make use of the fact that the standard implementation of the paradigms we rely on includes both forced and free choices to test this hypothesis.
[…]
GENERAL DISCUSSION
This study shows that decision inertia plays a role in human decision making under risk and investigates the underlying processes. We find a significant tendency to repeat previous choices in decision making with monetary feedback. Specifically, we found evidence for the existence of decision inertia in Study 1 and in decision situations without conflict with reinforcement in Study 2. In contrast, in the Left-Urn situations in Study 2, where reinforcement conflicts with Bayesian updating, we only found an effect of decision inertia after autonomous choices. We conclude that decision inertia seems to be subtle and easily overshadowed by stronger processes as e.g., reinforcement learning.
We hypothesized that decision inertia would be positively associated with PFC. The regression analysis confirms this hypothesis, indicating that the tendency to repeat past choices is a relevant part of the need to be consistent. This finding agrees with those of Pitz (1969), who showed that the inertia effect in opinion revision results from a psychological commitment to one’s initial judgments. It is not consistent with the results ofZhang et al. (2014, Study 2b), who found no relation between repetition of earlier decisions and PFC scores. However, Zhang et al. (2014) targeted unethical decisions and hence their setting is hard to compare to ours. The moral framing of the decisions in that work might have interacted with the hypothesized need for consistency. Our results for free vs. forced draws provide further evidence that decision inertia might (at least partly) be based on a mechanism of consistency-seeking. Both of our studies suggest that the effect of decision inertia might vary according to the type of first-draw decisions. The results of the regression analyses confirm this idea, indicating that decision inertia is significantly stronger in autonomous choices than in required ones. Since one would assume that a psychological desire to be consistent with one’s own decisions is stronger for self-selected compared to assigned decisions, this result further supports an interpretation of decision inertia as a facet of consistency-seeking.
Our results are also in agreement with the reinforcement learning literature (e.g., Schönberg et al., 2007; Gershman et al., 2009; Wimmer et al., 2012) which has pointed out the importance of perseveration as an additional factor. A direct comparison is of course difficult, because in our paradigm success probabilities are explicitly given (and priors are reset after every round), while in the quoted works they are discovered through experience. However, the basic messages are similar. As in those previous reports, we find that the mere repetition of previous choices plays a role even when behavior is mostly determined by the interaction of reinforcement and normative goals. In that sense, we confirm (in a different setting) that incorporating perseveration into models of reinforcement learning can improve our understanding of how errors occur.
In conclusion, we find clear evidence for the existence of decision inertia in incentivized decision making. Our study sheds light on the process underlying decision inertia, by showing that this behavioral tendency is positively associated with an individual’s preference for consistency, and that the effect of decision inertia is stronger in voluntary choices than in required choices.