Combating Confirmation Bias

Ah schemas. We’ve discussed them earlier in the book, but we need to take a closer look, as they are a major component in confirmation bias.

Imagine you want to offer your friends a pitted, pithy tree-fruit encased in red, edible skin. What do you do? You ask them if they want an apple. In this case, apple is your schema for a pithy, red fruit. Our brains rely on such schemas to make sense of the world. Without schemas, the complexity of reality would overwhelm us. According to

A schema is a cognitive framework or concept that helps organize and interpret information. Schemas can be useful because they allow us to take shortcuts in interpreting the vast amount of information that is available in our environment.

As you can see, schemas are powerful. They allow us to communicate efficiently, and at high levels of abstraction, all in one word. With schemas, we easily convey complex concepts without getting bogged down in details. Schemas play a key role in how we think.

But there’s a dark side. What we gain in flexibility and lucidity, we lose in objectivity. Here’s the rest of the definition:

… However, these mental frameworks also cause us to exclude pertinent information to focus instead only on things that confirm our pre-existing beliefs and ideas. Schemas can contribute to stereotypes and make it difficult to retain new information that does not conform to our established ideas about the world.

Uh oh. This, my friends, is confirmation bias. Tackling confirmation bias is by far the most difficult of the ‘aligning tricks’ to master. Because of schemas, we hear what we expect to hear. When we have a strong belief system, we hear what we want to hear. None of us are exempt. We’re all are guilty of cherry picking facts that support our preconceived ideas. It’s simply how our minds work. We all do it, whether aware of it or not.

Some of us, however, seek confirming data purposefully. We ask loaded questions to prompt answers we want to hear. The hallmarks of these ‘fishing expeditions,’ are, closed ended questions that steer customers toward specific answers – answers that might be used to justify a pet project, for instance. This behavior is unprofessional and intellectually dishonest. People who ask such questions are contemptuous of the customer, the truth, and the team.

Remember Cunningham bombs? When deployed, you sacrifice your dignity in exchange for a bit of information. The beauty of the Cunningham bomb is that the information you gain is generally unbiased, truthful, and useful. Contrast that with casting out a line on a fishing expedition with a closed-ended question. In this case, you sacrifice your integrity for biased, and often erroneous information.

Take, for example, the particularly egregious example of William J. Casey. Mr. Casey was Ronald Reagan’s campaign manager, and then, CIA director.

Before Reagan took office, Casey read The Terror Network, by Claire Sterling. This book connected the dots between various news articles that, taken together, seemed to link the USSR to global terrorist organizations. These revelations outraged Mr. Casey, and he shared the book with friends and colleagues.

Soon after his swearing-in ceremony, he called a meeting with Melvin A. Goodman, then CIA Division Chief and Head of Soviet Affairs. Casey asked Mr. Goodman to corroborate Ms. Sterling’s facts and conclusions. Mr. Goodman said he couldn’t do that. Casey demanded to know why not. In an interview with Adam Curtis, Mr. Goodman explained.

MELVIN GOODMAN, Head of Soviet Affairs CIA, 1976-87: “… when we looked through the book, (The Terror Network), we found very clear episodes where CIA black propaganda–clandestine information that was designed under a covert action plan to be planted in European newspapers–were picked up and put in this book. A lot of it was made up. It was made up out of whole cloth.”

INTERVIEWER (off-camera): “You told (William Casey) this?”

GOODMAN: “We told him that, point blank. And we even had the operations people to tell Bill Casey this. I thought maybe this might have an impact, but all of us were dismissed. Casey had made up his mind. He knew the Soviets were involved in terrorism, so there was nothing we could tell him to disabuse him. Lies became reality.”

Voice Over: “In the end, Casey found a university professor who described himself as a terror expert, and he produced a dossier that confirmed that the hidden terror network did, in fact, exist.”

This instance of confirmation bias is alarming. This wasn’t a simple case of subconsciously missing a few facts. Mr. Casey’s staff outright told him that they themselves had fabricated the evidence. He had to do some convoluted and disingenuous mental gymnastics to justify what he did next. He purposefully disregarded the evidence before him, and sought out alternate ‘facts’ to support his steadfast view that Moscow was directing global terror. Tragically, this ‘evidence’ contributed to Ronald Reagan’s decision to fight covert wars. American troops fought against an invisible, and largely imagined Soviet threat. It cost lives, resources, and America’s integrity in the eyes of the world.

Unfortunately, the case of William J Casey is hardly unique. People in similar positions of power and responsibility have acted in the same way. The Dulles brothers, who came to power decades before, behaved at least as irrationally. Stephen Kinzer documents many such incidents in his book, The Brothers. He wrote:

“both (Dulles brothers) suffered acutely from … confirmation bias – the tendency to reject discordant information. When their own envoys advised them to tolerate Mossadegh, and Arbenz, or to accept neutralist regimes in Indonesia and Laos, they could not hear. Instead they replaced the envoys with others who gave them the reports they wanted.”

Sound familiar? It gets worse. He continued,

“An American scientist, Ole Holsti, studied the way Foster made decisions, and found that he dealt with ‘discrepant information’ by ‘discrediting the source of the new information; reinterpreting the new information so as to be consistent with his belief system; [or] searching for other information consistent with preexisting attitudes.”

History is riddled with such examples.

The stakes involved in aligning with customers are hardly so high. Over time, however, if a culture values justifying beliefs over learning the truth, if people regularly discount disconfirming information, chances increase that the wrong product might be built, market share could be ceded, and profits forgone.

You’d think that leaders and planners would be rational stewards of their roles. You’d expect them to gather all relevant facts, and evaluate them dispassionately. To be sure, many do. Those rare people have faced up to their own prejudices, and come up with techniques to maintain some measure of immunity to confirmation bias.

Two people come to mind. John Costello and Richard Cliff. They were Vice Presidents of Engineering at Altera Corporation. They attended my customer meetings regularly. They would listen to customers with full interest. Their questions suggested they sought information that bucked their personal views. John might ask a question like, “You stated that you need x feature achieve y, however, my understanding is that x is usually used to achieve z. Could you please elaborate on how using x to achieve y will get you closer to your goals?” Open ended, not loaded, genuinely curious. Awesome.

I remember sitting next to Richard in one customer meeting when he poked me in the arm. He leaned in and whispered that one of my colleagues had just asked a loaded, leading question, and maybe I should correct his behavior. He said that he couldn’t trust information collected in that manner. Double awesome.

Both of these gentlemen had profound respect for the customers. They’d never go on fishing expeditions to manipulate the customer into saying something they wanted to hear. They had a job to do, and that was to build the right products for a given market at the right cost structure. Beyond that, they never seemed to have an agenda. It was evident that at some point in their lives, each of them had learned how to minimize confirmation bias. Whether they worked at it, or it came naturally, they were examples to follow.

And You Can Too!

As you now are all too aware, schemas are an integral part of cognition. So how does one overcome them? It’s kind of like someone with an eating addiction. They can’t simply not eat, like an alcoholic can do with drinking. That’s how it is with schemas, you can’t stop calling an apple an apple. I mean, you could call it a Gala, but, well, that’s still employing a schema. Like a person with an eating disorder, the healthy course of action is to acknowledge the challenge, and nurture new habits. The condition can’t be cured, but it can be managed.

A marketing professor, Mohanbir Sawhney, said, “To gain customer insights, we must understand that we are prisoners of what we know and what we believe.” The bars of this prison are our schemas. But there’s a way to unlock the gate.

Awareness of confirmation bias is a start, but it’s insufficient. Stamping out confirmation bias takes conscious effort, vigilance, well-established processes, and disciplined habits. If the director of an intelligence agency can succumb to confirmation bias so flagrantly, so too, can you. I advise you take action now.

Here’s the good news: if you’ve read the whole book up to here, you’re already half way down the path to managing confirmation bias. Most of the tips that counteract bias, are the very same techniques you’ve already learned in The Art of the Question, The Art of Taking Notes, and the Art of Listening.

Take verbatim notes. Write down everything the customer says. If you write down every word, it’s harder to then, later, willfully ignore or misinterpret what’s been said. You can still do it, but it’s more obvious when you do.

Ask ‘Active Listening’ Questions (ALQs). Ask open ended, customer focused questions that don’t ‘lead’ to particular answers you want to hear, but instead get the customer telling their story from their perspective, freely. Use restatement until the customer indicates they’re done. Make sure you get the full answer.

Write down questions prior to asking them. Until asking ALQs becomes second-nature, writing down complete questions makes you less likely to ask leading questions. This isn’t always possible, but it’s worth trying, especially when you’re learning how to ask ALQs.

Be comfortable not knowing the answer. Practice saying, “I don’t know.” We often want to resolve the unknown, or save a situation with an answer. It’s okay to not know what the answer is. Double down with uncovering the customer’s goals and challenges. Don’t worry about tying up loose ends.

Hold a post-customer-alignment meeting. Discuss what was heard and what it meant. Hear various interpretations and discuss them openly. Write down the team’s conclusion, and document alternate interpretations. Consider assigning some people to argue as devil’s advocates. These contrarians should challenge assumptions, premises and conclusions.

Publish meeting summaries and raw notes. The minutes and takeaways from every alignment meeting should be accessible to everyone involved in planning products. Invite questions and alternate interpretations. Make sure this data complies with customer confidentiality rules and procedures.

Bounce the meeting notes off the customers. Within a few days of the alignment meeting, send the customers your collated and verbatim notes along with a list of action items. Ask for corrections to the record or for missing information. Confirm the veracity of the record. No one can spot glaring errors, omissions, and misinterpretations like the customer can.

Create a culture of truth seekers. If you’re a leader of the group, go out of your way to appreciate the comments made by devil’s advocates in the follow up meeting. Demonstrate comfort and calm while holding two incompatible views at once. For example, acknowledge that many customers only care about cost, while many other customers care about performance. These cases can coexist.

Actively seek out disconfirming data. If a customer makes a statement that is contrary to conventional wisdom, ask follow up questions to explore their comment further. This is easy to do when a customer says something that is obviously contrary. Those comments can be so jarring, that missing them is impossible. Less obvious are comments that are only slightly off. Be on the lookout for those, and, when the time is right, ask for more details.

Appreciate the Contrarians
If you’re a leader in the organization, go out of your way, during the follow up meeting to compliment people who asked the customer about comments that were contrary to the conventional wisdom within your organization.

Keep in mind that, often, customer comments seem to be inline with your pre-conceived biases, when in fact, upon examination, they differ slightly. These are the comments that often go unexamined. As much as you tell yourself to keep an eye out for disconfirming comments, it is the slightly different information that will slip through. The only way to catch these comments is by taking verbatim notes, and as insurance, having both a meeting and a session scribe recording the meeting at the same time. Taking verbatim notes is critical.

Meet with customers who prefer your competition.
Talk to customers who hate you. Strive to understand what they’re trying to achieve, what their goals are, and in what ways they expect your competitors will help them make progress. Get them on a roll, espousing their angst! Again, double down on uncovering comments that run counter to your beliefs. Ask for goals and challenges with an open heart and mind. Whatever you do, do not disparage the competition. That’ll only make you look petty.

Ask deliberately stupid or naive questions. Disregard everything you know, and see if you can get the customer to explain even simple concepts. Employ Cunningham’s Law (explained earlier in the book). You may find the customer’s view on the world is completely different from your own. Record what they say.

Test yourself. Remember the video (discussed in a previous chapter)where Officer Bruch used discomfort with silence to get suspects to start talking? As an optional exercise, I encourage you to watch the entire video. It’s a treasure trove of examples of confirmation bias. How many instances can you count? Can you think of alternate approaches that the detectives or prosecutors could have taken in order to minimize confirmation bias? A link to my tally is in the appendix.

Adopt the mindset of a scientist / Don’t get married to your plans.
The act of planning products should be based on inductive reasoning, where you start with a set of objective observations and facts (that are future focused, Gretzky), and then come to a conclusion based on those. That conclusion is assumed good until observations or facts come along that disprove it. This is in contrast to deductive reasoning, whereby a set of facts leads to a conclusion that is, itself, a provable fact. There’s no way to be absolutely sure that a product plan will perfectly address a given market. New information is always coming in. Its everyone’s job to be on the lookout for new, disconfirming information at all times. They don’t need to bugle horn them to the entire company, but they should be discussed among the key members of the product development team.

Product plans are like theories that, at any time, can be supported or disproved with new evidence. If you’ve spent months creating a particular product plan, it only makes sense that you might become attached to it. You’ll want to support it with as much evidence as you can find. Careful. This is when confirmation bias flourishes. Take care to wear your scientist’s lab coat and be on the look out for new, disconfirming evidence.

We called this “getting married” to our plans. Resist! Don’t be a bride or groom, be a scientist who’s equally happy to find evidence that disproves a theory, as she is to find evidence that supports it. I know, it’s hard. The product you’ve been working on can feel like your baby. Why would you want to destroy your baby? The trick is to not think of it as your baby, but as a scientific theory – a theory that you’ll be lauded for disproving as much as you would for proving. If the leaders of the group are on the lookout for such clear eyed analysis, and praise it, you’re going to be fine.

This is the kind of thinking it takes to overcome confirmation bias. It requires focus, maturity, and strength.

Be fearless.
An occasional complaint I heard from colleagues was that we were blind to what our customers were saying because we operated from a place of fear, some more than others. The idea being that people were reluctant to propose risky changes to the product architecture plans because failure could result in them losing their bonuses or worse, their jobs. I’m not sure I was so cynical, but I could relate to operating from a place where retaining the status quo felt more comfortable than taking risks. The thing that helped us most, was to formally establish the goals of CA meetings: to fully understand the goals and challenges customers were facing. We strove to divorce ourselves completely from any personal wishes, at least for a few hours.

We allowed ourselves permission to understand the viewpoints of customers without feeling like we had to act on what they said. That’s the key. I’m not saying to not act on customer feedback. I’m saying you need to understand that you are not obliged to act on all customer feedback. That can be very freeing. Use that understanding to cultivate genuine interest for truly learning what the customer is trying to do.

If you can’t beat schemas, join them.
It’s not enough to simply know the definition of a schema. You need to know what your particular schemas are. In fact, I’d urge you to document the latest schemas. What are the goals and challenges of your typical customer in the automotive control systems end market? What are their preferences? How about the cloud storage market? What about the network infrastructure customers? Do the same for all of your end markets or end applications. Put together your best understanding of each type of customer. Document, and formally evolve your schemas with each CA meeting. Then, go into meetings and listen for comments that don’t fall in line. Probe them, and understand them.

Talk to many customers
There’s a tendency to believe that one customer represents all customers. This was Herb’s mistake in the Simpson’s episode, Oh Brother, Where Art Thou? (as discussed in a previous chapter).

When people extrapolate from a single data point, they can conclude anything. This can cause real damage as half cocked understandings are expressed with confidence and authority. It’s better to require that people attend a couple dozen CA meetings. Require them to contribute to the corporate schemas before materially participating in product planning activities. More on this in the chapter, Nineteen or None, which provides further insights on how many customers to talk to, and how to put the schemas together.

The Confirmation Bias Paradox.
If you seek out disconfirming information, meet with customers who don’t conform to your idea of what’s typical, and hash out alternative theories of what customers are truly trying to do. You may find your head spinning, as it can become too much information. Your job may already be confusing enough, and here I am, asking you to complicate things further by doubling down on understanding data that doesn’t jive with reality as you understand it. Sorry. Do it anyway. Here’s why.

Imagine your boss asks for your report on what should be done to address a particular market. You’re swimming in ambiguity, and find it impossible to give clear direction. How do you tease out a simple course of action from this jumble of qualified data points and tepid conclusions? Do you write up a memo that outlines various theories of what’s going on? Probably not. That’s the last thing your boss wants. Your boss is getting these reports from your colleagues as well, and has to make sense of them all. Ambiguous reports from all sectors could drive your boss mad. You think you have it bad? Imagine your boss’s dilemma. Indeed, this is why you’ve been asked to be decisive and keep your recommendations clear and simple.

I’ve seen this in action. People who tended to rise up the ranks were often people who proved themselves to be decisive authorities on specific subjects. They would confidently declare what course of action needed to be pursued. When proven right, they would be promoted. Usually these were highly qualified, and knowledgeable people. However, sometimes these were the very same people who got married to their ideas. And why wouldn’t they? They’d been rewarded handsomely for such ideas.

The problem was, that if they hadn’t developed habits of checking their assumptions, or of seeking out disconfirming information, they could become blind to changes in the market. Thus oblivious, they got further married to their ideas. Their bosses might love this, as their reports and suggestions came across as clear and confident. However, in time, they became victims of their own confirmation bias. Back to individual contributor they’d go.

This paradox – where one strives to dig up information that challenges assumptions, all while being expected to make decisive recommendations based on a sharp understanding of the situation – has a solution. It’s this: talk to more customers. And then talk to more customers. Keep talking to them, and approach each one with a curious mindset, with ALQs at the ready. Prepare to be confused. Get comfortable with it. Be patient, and wait for answers to emerge. Then be ready for those answers to be upturned by new information. This isn’t a magic bullet. It takes time, patience, focus, and hard work.

Consider sleeping on a bed of nails. That’s about the level of discomfort a truth-seeker needs to cultivate in order to overcome confirmation bias.

Confirmation Bias Summary.
It’s impossible to eradicate confirmation bias, as it’s built into how our brains process information. However, it can be minimized. The first step is to acknowledge that it exists, and understand how it works. Then, review your premises, and check the veracity of your observations and facts. Keep your eyes open for new information, especially anything that challenges conventional wisdom. Think of plans dispassionately, as scientific theories, rather than sweet cuddly babies. Use the techniques listed above to cultivate such a mindset.

When you ask ALQs, take verbatim notes, and listen actively. If you do all this, you’re well on your way to overcoming this insidious bias. The rest of the process is simply having an open mindset and actively paying attention. You can do it.