Posts Tagged ‘Heuristics’

A classic view regarding decision-making holds that attention serves foremost to acquire the information most relevant and important for choosing between alternatives. Thereby the role of attention is largely a passive one. However, an alternative view that is gaining traction in recent years, especially due to the help of eye tracking research, argues that attention plays a more active role in decision processes, influencing the construction of decisions.

This is a key message delivered by Orquin and Mueller Loose (2013) in their review on the role of attention in decision-making, as can be learnt from tracking of eye movements and subsequent fixations [1]. The approach taken by the researchers, however, is less usual: They do not constrain themselves concretely to the domain of decision-making; instead, they start their review and analysis of evidence from theories or models of tasks similar or related to decision-making (e.g., perception, information processing, visual search, working memory, top-down and bottom-up processes, problem solving).  Then they try to project how the functions of attention in such tasks may project to or be expressed in decision processes.

Furthermore, Orquin and Mueller Loose examine the extent to which the evidence coincides with four alternative theories and associated models of decision-making (i.e., whether empirical evidence substantiates or refutes assumptions or conclusions in each theory). They review evidence from previous research on similar or related tasks that could also be traced specifically in decision tasks, based on eye tracking in decision-making research, and evaluate this evidence in the context of the alternative decision-making theories.

The theories and related models considered are: (1) rational models; (2) bounded rationality models; (3) evidence accumulation models (e.g., the attention drift diffusion model [aDDM] posits that a decision-maker accumulates evidence in favour of the alternative being fixated upon at a given time); and (4) parallel constraint satisfaction models (a type of dual process, neural network model based on the conception of System 1’s fast and intuitive thinking [first stage] and System 2’s slow and deliberate thinking [second stage]). Rational models as well as bounded rationality models more explicitly contend that the role of attention is simply to capture the information needed for making a decision. ‘Strong’ rational models hold that all relevant, available information about choice alternatives would be attended to and taken into account, whereas ‘relaxed’ rational models allow for the possibility of nonattendance to some of the information (e.g., attributes or object [product] features). Bounded rationality models suggest that information is acquired just as required by the decision rules applied. The two other categories of models are more flexible in regard to how information is acquired and used, and its effect on the decision process and outcome. However, the authors argue that all four theories are found to be in deficit to a smaller or larger degree in their consideration of the role and function of attention in decision processes, having at least some of their assumptions being rejected by the evidence evaluated.

Selected insights drawn from the review of Orquin and Mueller Loose are enlisted here only briefly to shed light on the significance of attention in consumer decision-making.

A crucial question in decision-making is how information enters the decision process and is being utilised in reaching a choice decision: information may be acquired through attention guided by a top-down (goal-driven) process, yet information may also be captured by a bottom-up (stimulus-based) attentional process. The entanglement of both types of processes when making a decision is a prime aspect in this domain and has multiple implications. A more efficient selection process may be driven by greater experience with a task (e.g., more important information cues have a higher probability of being fixated on) and increased expertise in comprehension of visualisations (e.g., more fixations to relevant areas, and inversely fewer fixations to irrelevant areas, requiring shorter fixation durations, and longer saccades [‘jumps’ between more distant elements of information in a scene]). The interaction between bottom-up and top-down processing can amplify attention capture and improve the visual acuity of objects perceived. Bottom-up attention in particular is likely to be influenced by the saliency of a visual stimulus; however, it may not take effect when the task demands on attention are high, wherein priority is given to top-down directives for attention. Decision-making research has shown that visually salient alternatives or attributes are more likely to capture attention and furthermore affect the decision in their favour.

An interplay occurs between working memory and ‘instant’ attention: As the load of information fixated becomes larger, more elements are passed to working memory, and information is accessed from there for processing; however, as the strain on working memory increases, consumers turn to re-fixating information elements and consider them instantly or just-in-time (i.e., fixations are thus used as external memory space). This type of interplay has been identified in tasks of problem solving. Toggling between working memory and fixations or re-fixations in decision tasks can be traced, for instance, in alternative comparisons. Greater demands imposed by information complexity and decision difficulty (due to greater similarity between alternatives) may require greater effort (operations) in acquiring and processing information, yet the process may be shortened on the other hand through learning.

  • Another area with interesting implication is processing of visual objects: Previous research has shown that visual objects are not encoded as complete representations (e.g., naturalistic product images) and the binding of features is highly selective. Thereof, encoding of particular features during an object-stimulus fixation may be goal-driven, and a re-fixation may be employed to refer just-in-time to specific object [product] features as needed in a decision task, thus saving on working memory capacity.

Consumers have a tendency to develop a bias during a decision task towards a favoured alternative. This alternative would get more fixations, and there is also a greater likelihood for the last alternative fixated to be the one chosen (put differently, consumers are likely to re-affirm the choice of their favourite alternative by re-fixating it just before making the decision). A desired or favoured attribute can also benefit from a similar effect by receiving more frequent attention (i.e., fixations). The authors point, however, to a difficulty in confirming evidence accumulation models: whether greater likelihood of a more fixated alternative to be chosen is due to its higher utility or greater exposure to it. They suggest a ‘soft’ model version in support for a greater effect of extended mere exposure leading to choice of an alternative. They add that a down-stream effect of attention from perception onto choice through a bottom-up process may play a role of gatekeeping the alternatives entering a consideration set. It is noted that a down-stream effect, arising from a bottom-up process, is clearly distinguishable from a utility effect, since the former is stimulus-driven and the latter is goal-driven.

Consistent with bounded rationality theory, heuristics shape patterns of attention, directed by the information that a heuristic calls for (e.g., by alternative or by attribute). Yet, eye-tracking studies conducted to trace the progression of decision processes could not corroborate the patterns of heuristics used as proposed in the literature. More formally, studies failed to substantiate the assumption that heuristics in use can be inferred from the patterns of attention recorded. Transitions of consumers between alternative-wise and attribute-wise rules during a decision task make inferences especially difficult. Not only decision rules influence what information is attended to, but information cues met with during the decision process can modify the course of the decision strategy applied — consider the potential effect that salient stimuli captured unexpectedly in a bottom-up manner can have on the progression of the decision strategy.

In summary, regarding the decision-making theories, Orquin and Mueller Loose conclude: (a) firmer support for the relaxed rational model over the strong model (nonattendance is linked to down-stream effects); (b) a two-way relationship between decision rules and attention, where both top-down and bottom-up processes drive attention; (c) the chosen alternative has a higher likelihood of fixations during the decision task and also of being the last alternative fixated — they find confirmation for a choice bias but offer a different interpretation of the function of evidence accumulated; (d) an advantage of the favoured alternatives or most important attributes in receiving greater attention, and advantage of salient alternatives receiving more attention and being more likely to be chosen (concerning dual process parallel constraint satisfaction models).

Following the review, I offer a few final comments below:

Orquin and Mueller Loose contribute an important and interesting perspective in the projection of the role of [visual] attention from similar or related tasks onto decision-making and choice. Moreover, relevance is increased because elements of the similar tasks are embedded in decision-making tasks. Nevertheless, we still need more research within the domain because there could be aspects specific or unique to decision-making (e.g., objectives or goals, structure and context) that should be specified. Insofar as attention is concerned, this call is in alignment with the conclusions of the authors. Furthermore, such research has to reflect real-world situations and locations where consumers practically make decisions.

In retail stores, consider for example the research by Chandon, Hutchinson, Bradlow, and Young (2009) on the trade-off between visual lift (stimulus-based) and brand equity (memory-based); this research combined eye tracking with scanner purchase data [2]. However, it is worth looking also into an alternative approach of video tracking as used by Hui, Huang, Suher, and Inman (2013) in their investigation of the relations between planned and unplanned considerations and actual purchases (video tracking was applied in parallel with path tracking)[3].

For tracing decision processes more generally, refer for example to a review and experiment with eye tracking (choice bias) by Glaholt and Reingold (2011)[4], but consider nonetheless the more critical view presented by Reisen, Hoffrage and Mast (2008) following their comparison of multiple methods of interactive process tracing (IAPT)[5]. Reisen and his colleagues were less convinced that tracking eye movements was superior to tracking mouse movements (MouseLab-Web) for identifying decision strategies while consumers are acquiring information (they warn of superfluous eye re-fixations and random meaningless fixations that occur while people are contemplating the options in their minds).


It should be noted that a large part of the research in this field, using eye-tracking measurement, is applied with concentrated displays of information on alternatives and their attributes. The most frequent and familiar format is information matrices (or boards), although in reality we may also encounter other graphic formats such as networks, trees, layered wheels, and more art-creative diagram illustrations. Truly, concentrated displays can be found in shelf displays in physical stores and also in screen displays online and in mobile apps (e.g., retailers’ online stores, manufacturers’ websites, comparison websites). However, on many occasions of decision tasks (e.g., durables, more expensive products), consumers acquire information through multiple sessions while constructing their decisions. That is, the decision process extends over time. In each session consumers may keep some information elements or cues for later processing and integration, or they may execute an interim stage in their decision strategy. If information is eventually integrated, consumers may utilise aides like paper notes and electronic spreadsheets, but they do not necessarily do so.

Orquin and Mueller Loose refer to effects arising from spatial dispersion of information elements in a visual display as relevant to eye tracking (i.e., distance length of saccades), but these studies do not account for temporal dispersion of information. Studies may need to bridge data from multiple sessions to accomplish a more comprehensive representation of some decision processes. Yet, smartphones today can help in closing somewhat the gap since they permit shoppers to acquire information in-store while checking more information from other sources on their smartphones — mobile devices of eye tracking may be used to capture this link.

Finally, eye tracking provides researchers with evidence about attention to stimuli and information cues, but it cannot tell them directly about other dimensions such as meaning of the information and valence. The importance of information to consumers can be implied from measures such as the frequency and duration of fixations, but other methods are needed to reveal additional dimensions, especially from the conscious perspective of consumers (vis-à-vis unconscious biometric techniques such as coding of facial expressions). An explicit method (Visual Impression Metrics) can be used, for example, to elicit statements by consumers as to what areas and objects in a visual display that they freely observe they like or dislike (or are neutral about); if applied in combination with eye tracking, it would enable to signify the valence of areas and objects consumers attend to (unconsciously) in a single session with no further probing.

The review of Orquin and Mueller Loose opens our eyes to the versatile ways in which [visual] attention may function during decision tasks: top-down and bottom-up processes working in tandem, toggling between fixations and memory, a two-way relation between decision strategies and visual attention, choice bias, and more. But foremost, we may learn from this review the dynamics of the role of attention during consumer decision-making.

Ron Ventura, Ph.D. (Marketing)


[1] Attention and Choice: A Review of Eye Movements in Decision Making; Jacob L. Orquin and Simone Mueller Loos, 2013; Acta Psychologica, 144, pp. 190-206

[2] Does In-Store Marketing Work? Effects of the Number and Position of Shelf Facings on Brand Attention and Evaluation at the Point of Purchase; Pierre Chandon, J. Wesley Hutchinson, Eric T. Bradlow, & Scott H. Young, 2009; Journal of Marketing, 73 (November), pp. 1-17

[3] Deconstructing the “First Moment of Truth”: Understanding Unplanned Consideration and Purchase Conversion Using In-Store Video Tracking; Sam K. Hui, Yanliu Huang, Jacob Suher, & J. Jeffrey Inman, 2013; Journal of Marketing Research, 50 (August), pp. 445-462.

[4] Eye Movement Monitoring as a Process Tracing Methodology in Decision Making Research; Mackenzie G. Glaholt and Eyal M. Reingold, 2011; Journal of Neuroscience, Psychology and Economics, 4 (2), pp. 125-146

[5] Identifying Decision Strategies in a Consumer Choice Situation; Nils Reisen, Ulrich Hoffrage, and Fred W. Mast, 2008; Judgment and Decision Making, 3 (8), pp. 641-658

Read Full Post »

On the shelf in front of you are two jars of strawberry jam; the label on one of them carries the name of a well-known and popular brand of food, on the other jar appears an unfamiliar name. At the same price, which one do you choose to buy? The majority of consumers are likely to go for the well-known brand. It is a fast and easy-to-use heuristic — choose the product of the familiar brand name as it predicts the product is more likely to be of high quality. This fast-and-frugal decision rule is most useful when a consumer is not familiar with any of the jams from personal experience.

However, well-known brand names can affect preferences even when consumers do have experience with both products, and the products are in fact identical. This is a key theme surrounding blind tests of taste. For example, subjects in an experiment were given peanut butter to taste from three jars. The same peanut butter was put in all jars, yet one jar carried a familiar brand name whereas the other two were displayed with “no-name” labels. After tasting, 75% chose the branded jar for tasting better than the peanut butter in the two other jars (chosen by 17% and 8% of subjects). The power of recognized brand names is even more striking when mixing the labels between higher quality and lower quality peanut butter products: 73% chose the jar carrying a name-brand even though it contained the lower quality product (experiments conducted by Hoyer and Brown, 1990, cited by Gigerenzer, 2007). Familiar brand names may create a bias in their favour (“brand names taste better” as put by Gigerenzer). Researchers aim to remove the bias when testing how the taste of food products is perceived, but this potential bias should not be ignored when predicting consumer choice. Nonetheless, at time of choice, relying on the recognized name can aid consumers make correct choices more often under uncertainty at relatively little effort.

Take another example of the practicality of heuristics from a different field (criminology): the problem at hand is to find where a serial criminal might be located based on information on sites of his or her suspected crimes, a problem known as “geographic profiling”. The common approach is to apply sophisticated statistical models that calculate the probabilities of locations that are spatially distributed. In comparison, researchers Snook, Taylor and Bennell (2004, cited by Gigerenzer and Gaissmaier, 2011) tested a more economic “circle heuristic”: the criminal would be located at the centre of a circle drawn through the two most distant sites of crime; this heuristic performed better than ten alternative statistical profiling strategies in predicting the locations of criminals. Gigerenzer and Gaissmaier give this heuristic as an example for one-reason decision-making: the strategy employs a single and rather simple informative cue as its guide yet can perform better than more complex information-rich strategies.

Gerd Gigerenzer, a professor of psychology and Director of the Centre for Adaptive Behaviour and Cognition at the Max Planck Institute for Human Development in Berlin, commends a more positive approach towards heuristics and their role in decision-making than the critical view that has been accepted in psychology since the 1970s. A seminal paper published by Gigerenzer together with Wolfgang Gaissmaier (2011) is illuminating and instructive. In their article the authors work towards establishing a comprehensive theory of heuristics, and especially describe methods that need to be applied to properly define their strategies, specify the information they utilise, and assess their performance.

  • A heuristic is defined concisely by Gigerenzer and Gaismmaier as “a strategy that ignores part of the information, with the goal of making decisions more quickly, frugally, and/or accurately than more complex methods” (p. 454).  The function of a heuristic can fundamentally be specified by three building blocks: search rules, stop rules, and decision rules.

Gigerenzer disagrees with the paradigm in psychology on decision-making, led by Kahneman and Tversky and their colleagues. that associates heuristics with human error and biases. The paradigm is concerned with mistakes people make in processing information and drawing inferences (e.g., because of memory and cognitive limitations). Heuristics are used as a way of simplifying complex decision problems but are marred by inaccuracies and biases that result in sub-optimized  decisions or irrational decision-making. Gigerenzer and Gaissmaier assert that heuristics are not “irrational” and show that they do not necessarily “cost” the decision-maker in incorrect or less accurate outcomes.

Gigerenzer has had further reservations about the interpretation of “bounded rationality”, conceived by Herbert Simon, as the source of such biases or fallacies in human decision-making; Gigerenzer and Selten (2001) argue that bounded rationality is not identified with the class of error and fallacies in judgement and decision-making often demonstrated empirically in experiments in psychology and behavioural economics. In their view, bounded rationality is not about the discrepancy between human reasoning and laws of probability and optimization, looking instead for alternative norms to direct human decisions and studying the actual behaviour of “minds and institutions”.

On similar grounds, Gigerenzer and Gaissmaier (2011) take a critical stand towards the trade-off between accuracy of different decision rules or strategies and the effort in employing them, formally defined and extensively studied by Payne, Bettman, and Johnson. In the important framework of adaptive behaviour of decision makers, the researchers share concepts on rules and strategies as decision tools that people select from their “toolbox” (mostly subconsciously)  to fit the characteristics of each decision problem. Gigerenzer and Gaissmaier question, however, the implication that people use heuristics as imperfect tools to save time and effort but inherently sacrifice in less accurate or sub-optimal outcomes. The critique seems to target the type and variety of rules considered in analysing the accuracy-effort trade-off. They are more favourably oriented towards the alternative view of ecological rationality which converges more closely with bounded rationality, that is, how well a behaviour (e.g., using a heuristic) is adapted to fit the structure of the environment.

Gigerenzer champions the principle that “less-can-be-more”. It challenges the common rational and logical belief that methods and models using more information (as predictors) inherently lead to more accurate predictions. In environments with much uncertainty, when using more information, and applying it in complex calculations, it does not necessarily lead the decision maker to better results. Research on formalised  heuristics and alternative statistical models has often revealled “less-is-more” effects: “There is an inverse-U-shaped relation between level of accuracy and amount of information, computation, or time” (Gigerenzer and Gaissmaier, 2011, p. 453). It means that at some point adding more of those resources stops improving our performance and furthermore harms it. Hence, there are cases where a decision based on one reason (such as the “circle heuristic”) can be a sufficient and even a more advantageous method to achieving our goals.

Let us return briefly to the example on top of consumer choice of familiar brands. In the case that one of two alternatives (e.g., strawberry jams) is recognized and the other is not, a Recognition heuristic can be applied. The formal heuristic states: infer that the recognized alternative has the higher value with respect to the criterion (cf. Gigerenzer and Gaissmaier, 2011). The criterion, as suggested earlier, may be overall quality or taste. The efficacy of the heuristic has been demonstrated in quiz-like experiments. In one of such experiments, for example, participants were given pairs of cities in Switzerland and were asked for each pair which city had the larger population — the recognition heuristic succeeded in inferring the larger city in 89% of pairs (the rule is less accurate in inferring distance from the country’s centre, just 54%). However, one can benefit from the recognition heuristic only when he or she is ignorant in part, that is, one is familiar with just some of the objects and not others (i.e., when asking non-Swiss residents about Swiss cities).

  • Apparently, a study on tennis plays in Wimbledon 2004 has shown that amateur players, for whom the recognition heuristic is more applicable than for tennis professionals, were more successful in choosing the winner of plays in the tournament (72%) than the professionals or tennis experts (66%-69%).

When one is able to recognize both alternative objects in a pair, he or she can apply a Fluency heuristic which instructs to infer that the alternative recognized faster has the higher value with respect to the criterion. When we need to trust our own recall ability in generating alternatives, a third heuristic is specified by Gigerenzer and Gaissmaier, Take-the-First heuristic: choose the first alternative that comes to mind.

In his book on intuition and gut feelings published a few years back, Gigerenzer (2007) explains and demonstrates in an eloquent manner and a friendly style why and how heuristics work to our benefit. It is less formal and scientific-laden than academic articles yet is thus accessible to a wider audience of readers. The theory on intuition, unconscious cognition and heuristics is delivered in a way that is comprehensible and interesting, supported by many examples such as illustrations of the rules and in related research findings.

The heuristics identified by Gigerenzer and his colleagues are not meant to be newly invented — they may actually seem familiar to many people in one form or another. The significance of the contribution is in collating those heuristics from different fields of life, classifying and formally describing them according to systemic criteria, putting them into context of a theory of heuristics. This step is also important for analysing and evaluating their performance against more complex “rational” models.

Ron Ventura, Ph.D. (Marketing)


Gerd Gigerenzer, 2007, Gut Feelings: The Intelligence of the Unconscious, London: Penguin Books

Gerd Gigerenzer and Wolfgang Gaissmaier, 2011, Heuristic Decision Making, Annual Review of Psychology, Vol. 62, pp. 451-482.

Gerd Gigerenzer and Reinhard Selten, 2001, Rethinking Rationality, in Bounded Rationality: The Adaptive Toolbox, G. Gigerenzer and R. Selten (editors), Cambridge, MA; London, UK: MIT Press

Read Full Post »