Beware Path Dependence

In workplace groups, cognitive diversity usually trumps IQ; but not unconditionally.  Research by complexity scientist Scott Page demonstrates that a team of diverse thinkers tackling a problem will outperform the smartest individual in a group. The cognitive biases and limitations of even the smartest individual are usually no match for a group of minds correcting and building on one another. But this impressive benefit of team thinking is completely lost if its members are unwilling to challenge each other.

The reluctance to express opposing views troubled psychologist Irving Janis 40 years ago. Janis coined the term ‘groupthink’ to describe the tendency for teams to move towards unanimous conclusions because of individual desires to conform. Groupthink is just one form of a broader obstacle to high-quality discussion: the pervasive and dominating phenomenon of path dependence.
 
Path dependence is the tendency for things—such as events, belief systems and personalities—to unfold in ways that are partially constrained by the parameters of their path, and it influences all aspects of our personal and collective lives.  Simply put, the past determines the approximate boundaries of the path that the future takes.

The paradigmatic example of path dependence is the QWERTY typing keyboard. Designed in the late 19th century, it placed the letters used in common combinations, such as ‘el’ and ‘at’ far enough apart that the typewriter keys would not jam when the keys were hit in rapid succession. This keyboard layout remains the standard today, notwithstanding how difficult it is to learn and how inefficient it is compared to alternative key arrangements. The first typewriters established typing habits that became so ingrained they have been too difficult to overturn; they set a path we have not deviated from since.

The history of thought itself is path dependent, with each thinker reacting to or building upon the ideas of his predecessors. This is a clear trend in the history of Philosophy, from Plato’s influence on Aristotle to Hume’s influence on Kant, and it is equally apparent in religious evolution, from Hinduism’s influence on Buddhism to Judaism’s influence on Christianity and Christianity’s influence on Islam. Science is extremely path-dependent, because most research and discovery is related to previously undertaken observation and experimentation.

Our personal beliefs are also highly path-dependent: children brought up in atheist households don’t typically become religious fundamentalists; our perspective of romance is deeply influenced by our early romantic experiences and our parental interactions; and we are born into particular cultural milieus that direct our prejudices, preferences and perspectives onto paths we tend to stay on.

In short, much of what we choose to do is highly dependent upon our prior choices and experiences. Our careers have unpredictable turns and twists, but there is a great deal of path dependence in how they unfold: police officers rarely switch paths to become, say, Philosophy professors, and physicists rarely change course to become advertising executives. While the specificity of particular paths is not predictable at the outset, their influence on future events is clear in hindsight: paths direct the systems that they shape.

Path dependence reflects both random and non-random elements. A young boy witnesses the burglary of his grandfather’s store and later decides to become a police officer, eventually working his way up to chief of police. The random starting point of the burglary set the man on a path that was not pre-determined; his path would have been different if he had not witnessed the crime. What we typically think of as randomness cancels out over time (as in the ratio of heads to tails, which approaches 50 per cent after numerous coin flips). But this kind of linear randomness does not accurately describe the kind of pattern that arises from self-reinforcing behaviour (the rich get richer, the famous become more famous.) In these non-linear cases, events are fueled by positive feedback loops that reinforce rather than cancel out. These feedback loops amplify the influence of the starting point, and path dependence characterizes this reinforcing behaviour.

Path dependence is not necessarily a bad thing: it’s just a fact of life, an example of how things work when a random event evolves in a way that is not purely random. But path dependence can be problematic when it surreptitiously limits the options available to us, including, and especially, how we think.

If we are aware of the paths that we gravitate to, we are less captive to them. Being cognizant of the mental paths that constrain us is crucial if our individual thinking is to be rigorous, and it is no less important for productive group discussions. Path-dependent thinking in the individual realm can be even more problematic in the collective: path dependence permeates every conversation, including board meetings, executive team meetings, strategic off-sites, jury deliberations, political agendas, etc.

The starting point of any discussion has a disproportionate influence on the path that the conversation takes. Subsequent comments advance from this starting point in ways that are barely perceptible to the participants. When a point of view is tabled, it can gain momentum through the support of a few and be carried down the path to a firm conclusion without ever facing sufficient challenges. Sometimes, ideas become ingrained not because they are explicitly endorsed, but because they are not explicitly critiqued. Legitimate expressions of concern can be swept away by the appearance of a majority support that may not really exist. The same problem arises when a discussion has gone on for a period of time: participants become weary and inclined to acquiesce to a conclusion, even when the complexity of the problem may require more poking and prodding.

Path dependence is not limited to the starting point of a discussion; it occurs even when a strong counter-position is itself not challenged, veering onto a different but equally constricted path. When thinking doesn’t stray from certain parameters, teams relinquish the potential for creativity and best-outcome results.

Figure One represents six hypothetical discussions on the same topic. The topic could be anything; for example, what price to bid for a strategic acquisition, or how much a CEO’s compensation should be contingent on company results. Of the infinite number of possible discussions on the topic, the paths of six are illustrated. The starting positions on the left are the initial viewpoints that are expressed: the recommendation of a company’s head of strategy to the CEO’s management team, for example, or the presentation by the head of human resources to the compensation committee of the board.

The path of the discussion moves from left to right. Each movement up or down represents a shift in thinking at the table; for example, to pay more or less for the acquisition, or to make the CEO’s bonus a greater or lesser portion of her compensation package. The ending positions are the groups’ concluding decisions. The shaded area represents ending points that are within the range of best (most productive) outcomes.

Six Conversations
Discussion A starts with a position that is endorsed by all participants, and it concludes in the same spot. Because it does not deviate from its starting position, it is highly path dependent. Discussions B and C share the same starting position, but B is quite path dependent, veering only slightly toward a better solution, whereas C represents a fulsome conversation covering a number of perspectives and ending in a range of good solutions. Discussion C reflects the highest quality conversation of the six because it resists path dependence to the greatest extent. D and F have different starting positions, but they conclude with identical solutions. E starts with a workable solution but is captive to a strong emerging path that directs it below the range of acceptable outcomes.

Path dependence is a problem when the wrong path dominates, taking the discussion to the wrong conclusion and bad decision. A good path can usually withstand challenge, as F did, but E failed to do so. A bad path won’t necessarily reveal itself unless a high-quality conversation ferrets it out.

The fact is, as individuals, we are cognitively lazy: we take mental shortcuts to sort out information and hurry to conclusions, because our brains evolved to sort through and respond to the barrage of sensory data we’re constantly exposed to. Groups suffer from the same cognitive vulnerabilities: reluctance to self-examine, cognitive laziness, pre-set biases, etc. Whereas individuals are hard-pressed to be aware of and overcome their biases and limitations, groups can be highly effective in both correcting each other and building on ideas. But the group benefit only accrues if discussion is not constrained by the limits of one path: participants must participate with a degree of independence from one another, so that the conversation is not dominated by a single stream of self-reinforcing thought.

Without independent thinking, everyone can easily gravitate to the same error. The stock market is a good example: markets tend to reflect the collective wisdom of all participants, but stocks can overshoot on the upside (unwarranted optimism) and on the downside (excessive pessimism). Price bubbles and depressions are the result of investors being influenced by each other, generating reinforcing feedback that extends euphoric or depressive sentiment, until the loops are finally broken and the market reverses direction, gravitating back to intrinsic values. Unlike stock markets, team discussions usually terminate in conclusions, so a lack of independence has no opportunity to self-correct. The termination points can be just as extreme, though, as stock market bubbles.

Harvard professor Cass Sunstein observes that teams often become emboldened by consensus and arrive at more definitive, extreme conclusions than the individuals’ going-in positions. What he calls ‘group polarization’ occurs when initial views are accentuated by discussion, generating momentum on a single path without counterbalancing influences. Sunstein points to two contributing factors in the origin of polarization: the problem of social influences, such as a desire to conform and the persuasion of passionate presenters, and the problem of ‘limited argument pools’, where information that challenges the prevailing argument is unavailable.

Sunstein concludes that avoiding polarization requires structured deliberation to ensure participants are exposed to alternative lines of thinking. His conclusion complements research by psychologist Charlan Nemeth, who focuses on how easily ‘majority thinking’ squeezes out ‘minority thinking’, notwithstanding the value that minority disagreement offers. Majority thinking tends to be confirmation driven, where rationale that supports preliminary conclusions builds momentum toward a converging perspective. Even when we are persuaded by minority dissent privately, our tendency is to publically support the majority view.

Nemeth challenges the accepted practice of free-flow brainstorming. She cites experiments showing that the ‘no idea is a bad idea’ form of brainstorming generates less creativity (fewer ideas) than brainstorming where participants are encouraged to challenge and debate each other. Critique improves the quality and quantity of the deliberative process; it both broadens and deepens discussion. Constructive argument tends to invoke more information and contemplate more alternatives.

Constructive dissent is the antidote to path dependence, and therefore a crucial component to team discussion. Dissent is necessary to ensure outcome quality because it both de-correlates the errors and biases of each individual, and deepens the quality of thinking in general. Dissent breaks the momentum of rigid paths by opening up new possibilities.

Dissent itself depends on two things: heterogeneity of thinking, and conversation to leverage that heterogeneity. Because our individual cognitive limitations mean that we are rarely our own most rigorous critics, the best solutions are frequently inaccessible to a single cognitively-constrained mind. Groups of cognitively-diverse thinkers can both correct each others’ mistakes and expand on others’ ideas, but only if dissent is permitted. However, it is easier to bring together people who have different backgrounds, expertise and approaches to problem-solving than it is to generate discussion among them that is not captive to a dominating path.

Competing arguments are the only way to deepen discussion, and as a result, fostering constructive dissent must be one of the main objectives of any team leader, no matter the arena: politics, business, academia, law, even religious institutions.

In closing, a note of caution: path-eroding dissent must be authentic to be effective. Nemeth is skeptical about the value of appointing a ‘devil’s advocate’ on a team, because contrived dissent is too easily dismissed by others. Experiments suggest that it not only fails to deepen the conversation; it also cements initial positions. Authentic and constructive dissent is crucial to realize the benefit of group thinking by avoiding groupthink: it pushes the flow of conversation outside a given path so that the line of thinking can be examined objectively and alternatives considered. Dissent fuels the dialectic process required to make group discussions productive and powerful.

Ted Cadsby, MBA, CFA, ICD.D, is a Corporate Director and the principal of TRC Consulting. He is the former executive vice-president of Retail Distribution at CIBC, and the author of three books, including Mind — the Gap: Smarter Decision-Making  in a Hypercomplex World, which will be published by BPS books in March 2014 .

[This article has been reprinted, with permission, from Rotman Management, the magazine of the University of Toronto's Rotman School of Management]