Mirjam Neelen & Paul A. Kirschner
“Life can only be understood backwards; but it must be lived forwards.”
Sanchez and Dunning (2018) recently published an article with the title ‘Overconfidence Among Beginners: Is a Little Learning a Dangerous Thing?’
This question is critical in the workplace context. There’s a tendency in organisations to make statements such as ‘everybody needs to be aware of topic X’ (where topic X is new technologies, new trends, new processes, and the like). The solution that’s usually chosen is providing the person / persons in question with topic-relevant content such as articles or videos or sometimes even a seminar. Note that all of these solutions have relatively little to do with learning – they’re only providing access to information; that’s all. In light of Sanchez and Dunning’s question, we need to ask ourselves: Is it wise (that is to say, is it productive) for an organisation to invest time and money in this type of solution? Is it possible that providing such information is actually risky (for) business? Why? People who, for example, engage with the information provided might believe they’re then knowledgeable about the topic and are also able to apply what they’ve been exposed to in their daily work, while that’s actually not the case. Also, they might try to apply the info they’ve been exposed to in their work tasks or discuss the topic with their clients, likely doing it wrong and making mistakes with all its concomitant consequences because they’re overconfident and unaware of how unknowledgeable they are. This is known as the Dunning-Kruger effect and we’ve blogged about it before. In short, the Dunning–Kruger effect is a cognitive bias wherein people of low ability suffer from illusory superiority, mistakenly assessing their cognitive ability as greater than it is.
Probably, many of you are familiar with the ‘four stages of competence’ model (originally ‘the Four Stages for Learning Any New Skill’ (Burch, 1970), see image below) which states that newbies are overconfident and make many errors (they’re unconsciously incompetent, aka Dunning-Kruger effect). As the newbie progresses (s)he moves onto a ‘conscious incompetence’ phase in which (s)he’s quite aware of what (s)he does or doesn’t know / can or can’t do. Next, the person becomes consciously competent, where (s)he’s able but still needs to put in quite some effort doing whatever it is (s)he needs to do. Finally, a person reaches the phase of being unconsciously competent or, proficient (in German this is known as having Fingerspitzengefühl; having it in one’s fingers).
Interestingly, Sanchez and Dunning come up with a different hypothesis. One that we feel makes a whole lot of sense…
The beginner’s bubble hypothesis
The authors propose a pattern of development which they call “the beginner’s bubble” (p. 11). What they suggest is that people do NOT begin at an unconscious incompetence level. Instead, they suggest that people approach a new task being quite cautious and unconfident in their decisions. Then, after a little learning, they quickly become overconfident (they enter the beginner’s bubble) before going through a correction phase in which performance improves and confidence levels flatten.
The beginner’s bubble 😊
Sanchez and Dunning focus on what they call ‘multicue probabilistic tasks’. These characterise the type of challenges that people face in work, namely tasks that have or give you some cues or hints as to how to proceed, but as an individual you need to make predictions about uncertain events. Let’s take as an example implementing a new technology within an office or a production process. The person charged with carrying out the task has access to content covering the technology (e.g., how it works and in what contexts it can provide value). However, deciding whether or not to make use of the technology and how to do it is, of course, not straightforward as there’s no ‘one right decision’ due to the many uncertain factors influencing the decision. In the context of a company’s choosing to ‘making people aware of topic X’ (here the technology), we run the risk that when dealing with a multicue probabilistic task, the person will base her or his decisions on what (s)he thinks (s)he knows about the topic, which again, might be a completely wrong perception of reality.
Sanchez and Dunning conducted six studies. The primary focus of the first four was on probabilistic learning. They examined whether beginner confidence and overconfidence arose in the specific pattern that they predicted (namely, unconfident at first and quickly overconfident after a little learning). In the fourth study, they also investigated whether enthusiastic theorising was at the basis of the pattern of confidence; that people, as soon as they have a little knowledge, rapidly form theories about how to best approach a task based on a too little data without realising that that’s what they’re doing. In studies 5a and 5b they switched from a lab environment to a real-world task (i.e., examining data on financial literacy across the life span) to see whether it followed the same pattern of subjective (confidence) and objective (accuracy) learning curves.
The beginner’s bubble: It takes little learning for overconfidence to develop
All of the studies confirmed that people are NOT overconfident when they’re absolute beginners. However, it only takes a little bit of experience for overinflated confidence to kick in. While this flattens over time, confidence always ‘outperforms’ accuracy. In other words, over time people develop more accurate components in their theories BUT they also continue to have components of error WITHOUT REALISING IT and are always more confident than accurate.
Sanchez and Dunning’s data suggest that a ‘boundary’ condition for the Dunning-Kruger effect might be that a person is an absolute beginner at a task or skill. This ‘condition’ only lasts a short time after which they become “some of the most vulnerable individuals to the Dunning-Kruger effect” (p. 24).
What to do to decrease the risks that come with the beginner’s bubble?
The philosopher R.G. Collingwood suggested the following:
While this sounds like a good starting point, we can do other things as well.
First, we can support people in becoming better self-directed learners (see our blogs here and here); although people might suffer from Dunning-Kruger effect here as well in assessing themselves as strong self-directed learners while they’re really not.
Also worth considering, is that ‘bombarding’ people in the workplace with content because ‘they need to be aware of topic X’, is NOT the way to go as it causes a risk to the business because people think they know and can use topic X).
If you think content is really necessary, you could intermittently offer ‘knowledge checks’ so that people can at least check what they actually DO understand based on the content they engaged with (and yes, of course those formative assessments need to be designed, including providing the necessary type of feedback). Also, it might be worth proving guidance with regard to what people are supposed to be doing with that supposed knowledge (i.e., the information provided). Show examples of people having ‘wrong’ conversations with clients and let people reflect on the ‘why’. Ask them questions, such as “Imagine your client would ask you about topic X, what could you give as an answer”?
Beware! Offering loads of content to people might do more harm than good. You’re actually investing money to achieve the Dunning Kruger effect and that’s without a doubt not only a stupid investment, but also a very dangerous thing!
Burch, N. (1970). The four stages for learning any new skill. Gordon Training International, CA.
Sanchez, C., & Dunning, D. (2018). Overconfidence among beginners: Is a little learning a dangerous thing? Journal of personality and social psychology, 114(1), 10.