Someone reads every report on AI. They follow the newsletters, attend the webinars, and bookmark the frameworks. Monday morning arrives, and they make the same calls the same way. Nothing changed. Yes, knowing more can make you less capable, but only when knowledge piles up faster than your ability to classify it, weigh it, and act on it. Knowledge is not the villain. Sequence is the point.
The short answer is that more input does not guarantee better output. Several lines of research point the same way.
A 2022 study in Judgment and Decision Making by Aßmann, Betsch, Lang, and Lindow found that even highly capable and motivated participants looked up twice the amount of information they needed. When the information conflicted, leaning on the less relevant pieces pulled decisions away from accuracy and produced worse outcomes. These were not careless people. They were the people you would want in the room.
An earlier line of work reached a similar conclusion in an organizational setting. A 2001 paper presented at an informing science conference reported that more factual information hurt individual decision makers’ processing efficiency and decision accuracy.
The British Psychological Society’s research digest adds a twist. In a study on causal information, participants with no personal experience of a topic benefited from extra causal information, with accuracy rising to 86.6%. Those with personal experience got worse. Their existing mental models shaped how they absorbed the new material. Samantha Kleinberg, a cognitive scientist at Stevens Institute of Technology, put the practical point this way: how information is tailored to what people already believe matters as much as how much of it there is.
So the honest answer to “can a person know too much?” is this. Volume alone is not the problem. Volume without structure is. And the more experienced the person, the more their existing mental models decide what the new material does.
There are three distinct ways more knowledge makes people less effective at work. They are worth separating, because each needs a different fix.
When someone faces a situation and holds a large pile of loosely sorted facts, the first job is not deciding. It is sorting. What kind of situation is this? Which facts apply? Which are noise? Without quick answers to those questions, the pile adds cognitive load, and load shows up as hesitation. People delay. They ask for one more report. They run one more analysis.
Peter Fiske, writing in Science, described this as the “Curse of Brains”: intellectually strong people sometimes analyze so much that they delay action, creating mental work and anxiety along the way.
Fiske also noted that once you have expertise, it is impossible to not know what you know. In approaching a problem, the expert tends to lean on what they know rather than consult others or try nontraditional paths. That is efficient when the problem looks like past problems. It is a trap when it does not.
Liz Wiseman, an organizational researcher, describes the flip side. In a 2016 BYU Magazine piece, she reported that people new to something hard and important, unencumbered by knowledge, see more possibilities, ask better questions, listen more, and mobilize other people’s expertise. Her conclusion is not that ignorance is a virtue. It is that questions and speed of learning count for as much as stored answers.
In teams, the knowledge exists, but it is spread unevenly and poorly connected. A 2016 Organization Science study by Reagans, Miron-Spektor, and Argote found that knowledge utilization and coordination each added much more to team performance when the other was present. When members could not coordinate, using knowledge actually undermined performance.
Nielsen Norman Group describes a related pattern called the common-knowledge effect: teams over-rely on what everyone already shares and under-weight the unique expertise one member holds. A 2011 field study of 177 teams in a multinational found that when everyone routes critical knowledge through one person, executive-rated team performance suffers, more so across diverse business units and when the knowledge is exploratory.
Here is the pattern we see underneath all of this.
Knowledge sits in memory and documents. Action happens in a moment, in a workflow, under time pressure. Between them sits a step most training skips: classification. Is this a routine case or an exception? Is the risk low or high? Does it fall inside my authority? Do I need more input, or do I already have enough?
Until someone can answer those questions quickly, additional knowledge has nowhere to land. It sits unsorted. That is the More Tangle in miniature: the belief that more knowledge automatically means more capability. We will take Tangles in full in a later piece. For now, one line is enough. More information adds load and hesitation when people cannot classify, decide, and act.
A simple sequence can carry the weight:
Treat it as adaptable, not mandatory. A nurse, a claims adjuster, and a field service lead will each compress or reorder these steps. The point is not a ritual. The point is that consultation comes after classification, not before it.
An assistant that produces a perfect analysis does not break this rule. It makes it more important. The better the intelligence available, the more the person’s classification and risk judgment decide whether it changes anything. We cover that gap in why AI insights often fail to change operational performance and in why decision capability sits at the center of AI and human collaboration.
If knowing is not the goal, what is? We define demonstrated capability as reliable decision performance under realistic conditions in real workflows.
Three phrases matter in that definition.
A quiz measures recall. A completion report measures attendance. Neither tells you whether a decision at the point of work went better. That is why completion numbers can climb while performance stays flat, and why leaders who are accountable for results, not course counts, grow impatient with them. The same logic runs through our piece on training effectiveness and decision quality.
Take a customer support organization that rolls out a new knowledge base and an AI assistant. Within a quarter, agents have access to more answers than ever. Escalations stay high. Handle times drift up. Nobody is short of information.
What is missing is a small set of shared rules. Which tickets can an agent resolve alone? Which need a second look? What signals mean a customer is about to churn rather than just ask a question? When agents lack those classification rules, they open more articles, hedge more answers, and escalate to be safe. The knowledge is there. The sorting is not.
This is an illustration of the pattern, not a customer result. But the research on teams supports its shape: knowledge needs coordination to help. And Linhardt and Salas argued in a 2023 Frontiers in Psychology article that expert teams with fluid membership widen the available knowledge but struggle to work together, so enduring impact depends on planning and reflection that integrate members cognitively and socially. For teams drawing on varied expertise, these innovation strategies for teams are one practical way into collaboration.
Reflection has its own limits. A 2018 Journal of Organizational Behavior study by Breugst and colleagues found that with reliable information, more information elaboration went with better team decisions at low levels of team reflection, and with worse ones at high levels. Reflection helps, but it needs to be matched to the task. More discussion is not automatically better either.
When a performance problem appears, the common reflex is to build training. Another course, another module, another library. It feels like action. But if the root issue is not a gap in what people know, more content only adds to the pile.
The better order starts earlier:
This is the shift we argue for in why best practices are not consistently followed: diagnose first, then decide how much to train. Harvard Business School’s Working Knowledge series makes a related point from the learning-organization side: organizations need to acquire, interpret, and then apply information by translating it into action. Garvin added that epiphanies alone do not make a learning organization. It takes a mindset of inquiry and experimentation plus a process for sharing what was learned so everyone can act on it.
Perspective
We do not think organizations know too much. We think they sequence it badly.
Most enablement starts at the end: a topic, a deadline, a request for a course. We start at the beginning. What does the work actually require people to decide? Where does the environment, not the person, get in the way? What does the organization’s own evidence say about where decisions go wrong?
That is why we built Cognistry as a system for diagnosis, evidence, and assurance, not a slide tool and not a course generator. The courses and simulations are real, and they matter. But they are the downstream result of knowing what capability to build and why.
The people you already employ hold a great deal of knowledge. A few clear classification rules, practiced in realistic situations, often unlock it. The goal is not less knowledge. It is knowledge that arrives in the right order, at the moment of the decision.
Brian
You do not need a new program to start. These moves apply to a team this week.
Pick the recurring decisions where errors are costly or slow. Write each in plain words. Vague categories cannot be classified.
For each situation, define two or three questions that sort it: routine or exception, reversible or not, inside or outside your authority. Keep them short enough to use under pressure.
Decide in advance which sources to consult for which class of situation. A low-risk routine case may need nothing. A high-risk exception may need a named expert. This limits the over-searching the research describes.
NN/G recommends promoting psychological safety, setting aside past opinions, and publicly recognizing members with domain expertise so unique knowledge gets shared. That counters the pull toward what everyone already knows. Our write-up on scaling expertise systems goes deeper on getting expert judgment out of a few heads.
Rules read on a page do not become habit. Decisions rehearsed with realistic pressure and partial information do. This is where practice simulations earn their place, after the diagnosis tells you which decisions to rehearse.
If your organization is deciding whether to build more training, we start with the question before the build. The Cognistry Platform covers the path from diagnosis to measured outcomes: Signal maps the capability signals the work requires from your own evidence, Forge structures the response, and Sim builds decision practice environments for operational judgment.
If you are accountable for team performance, not just completion, see how the Cognistry Platform works. Operations leaders can start with our page for operations teams.
Not in the sense that knowledge itself harms. But unsorted information can slow decisions and reduce accuracy. In one study, capable participants looked up twice the information they needed, and conflicting data pulled them off course. The fix is structure, not ignorance.
Only when they can apply it. Knowing facts and making reliable decisions in real workflows are different things. Experience can even change how new information is processed: in one study, people with personal experience of a topic got less accurate with extra causal information, while novices improved.
Experts cannot un-know what they know. Peter Fiske observed that experts tend to rely on their own knowledge and experience rather than consult others or explore new paths. Asking better questions and drawing on other people’s expertise helps offset that.
Coordinate before you collect. Teams that cannot coordinate their efforts see knowledge use hurt performance. Clear decision roles, shared classification rules, and reflection matched to the task keep unique expertise from being lost.
Not before a diagnosis. First work out what capability the work requires and whether the gap sits in the person or the environment. Then decide whether learning is the right enablement play, and ground any build in your own evidence.