Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Share Help Cite Discuss in SciProfiles

Open AccessArticle

Peer-Review Record

The Illusion of Control: How Knowledge and Expertise Misclassify Uncertainty as Risk

Risks 2025, 13(10), 188; https://doi.org/10.3390/risks13100188

by Alessio Faccia^1,*

, Pythagoras Petratos² and Francesco Manni³

Reviewer 1:

Josef Oehmen

Reviewer 2:

Vicki M. Bier

Risks 2025, 13(10), 188; https://doi.org/10.3390/risks13100188

Submission received: 8 October 2024 / Revised: 1 September 2025 / Accepted: 24 September 2025 / Published: 1 October 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

thank you for an interesting paper. I am offering a number of suggestions for your consideration in the following. I very much like the intend behind your paper, but I believe it needs significant additional work.

Abstract / General Remarks:

- Is "uncertainty" only relevant for unknown or not precisely known probabilities of occurence? What about the magnitude of the impact? If we view risks as probability distributions across negative outcomes, the two become the same. In the abstract, you make it sound like "uncertainty" only applies to probabilities of occurence, but NOT to a lack of knowledge of possible / plausible consequences.

- You frame risk as "probability of occurence of specific event" x " impact of specific event". As discussed above, that is a very "basic" model of risk. If you want to limit your paper to that definition, you need to make that very explicit.

- "valuable at assessing and managing risks" - please be more careful with your use of technical terms. Risk management is (at least) identification, assessment and mitigation - and you can make that as complex as your want. Talking of "assessment and management" makes no sense. And what about identification? Is expertise not useful there as well?

- You are trying to squeeze two contributions into one paper: introducing a novel sense making framework (and evaluating its efficacy), as well as introducing "process-based accountability". My default recommendation is: Make it two papers that then each can very clearly focus on the one point it is making. It is then much easier for you to have a targeted introduction and literature review. It is OK to publish two papers out of one research study.

- Abstract, line 23, "The study also highlights...": You are making the same points three times in a row - line 23-30 should be two lines

Introduction:

- Line 33 - 105: I did not count how many assertions you make, I would estimate around 25-30. That requires 25-30 literature references. I counted the current literature references: 0 .

- I am sympathetic to the argument you are making, but you offer nothing to substantiate your opinion. Its the "introduction" to a publication, not "Our personal reflections on the current state of affairs, based on our opinions."

- Line 72: Here you start the "second story" that I believe belongs into a second paper

Literature Review:

- I cannot help but think that your literature review remains somewhat superficial. I agree that you are identifying some of the seminal works, but the concepts of risk, uncertainty, ignorance (and sometimes myopia) have been discussed at much more intensity and depth than your literature review suggests. I strongly suggest to significantly expand the literature review, so that the readers are convinced that you are building on a strong foundation of existing knowledge and conversations in the field.

- The same applies in the continuation of your section in 2.4 and 2.5. The role of experts, expertise, calibration of expert judgement, elicitation of expert judgement etc. are discussed at much greater depth in the literature. The same is true of the discussion of cognitive biases, inside and outside of risk management.

- 2.7, your critique of ISO31000. This has also been discussed in the literature. The section is mostly a summary of your reasoned opinions, not a literature review.

- 2.8, again, a narrow selection of a small slice of the literature that is used extensively - to the point of beginning to seem like a summary of a handful of papers and books - thoughout the section and sub-section. Your job is not to re-tell a handful of papers.

- 2.9 / 2.10 - as I said before, this is a second paper.

Methodology:

- You mention "framework generation through dichtomoty matching" , "generation of conceptual frameworks", "analytical process involving matching and combining" as research methods. All of these require references regarding their origin and validity. The references you provide focus on the content of your analysis, not the method of your analysis and theory and/or framework building.

- 3.2: The origins of the rubrics is not clear to me. Why these? Why are they complete?

- 3.3: How did you conduct your case study? Last time I counted, I think it was 12 different types of case studies that I came across, and I do not think that is an exhaustive list. Each with their own methodological approach and validity criteria. What kind of case study are we talking about here?

- 3.4: How is the refinement done? I assume this also relates to the previous theory development? What theory building framework are you building on?

Results and Framework:

- The results seem to fall from the sky. What theory building happened here? What is mapped on what? Why is this valid / how was it validated? It is hard to see more than artistic intuition behind the framework, as you do not articulate your theory building foundation.

- 4.2: Quadrant description: You are describing the content of quadrants based on... what? Has any literature been mapped into the quadrants? Are you hypothesising what may be in the quadrants?

- 4.3: Rubric development: Again, they seem to fall from the sky, not "being developed". Again, for theory development, there is not a single literature reference here, or reference to some empirical work (interviews, observations etc.)

- 4.4: These are "illustrative examples without empirical data or references" at best. These are NOT case studies.

- the second 4.4 of the paper (page 28): It is hard to give credibility to improvements derived from illustrative examples. There is no empirical testing of your framework at this point, so this section is not credible. I strongly suggest you do test your framework with practitioners (say, in a handful of focus groups) to collect empirical data. You can then improve your literature-derived framework based on that empirical feedback. I suggest to skip the illustrative examples (or maybe use one as part of a first-pass sanity check after the theory development, but before the empirical testing).

- 4.5 I struggle to see how this is a different step from (the second) 4.4. Again, this lacks credibility beyond "we followed our intuition", as there is no new empirical evidence introduced. Also, see my standard comment regarding accountability mechanisms - that is a different paper.

- The two sections 5 (conclusions and study novelty...) need to be thoroughly revised based on an updated manuscript.

Author Response

General Comments

We would like to thank both the editors and in particular the reviews for their time and substantial effort to provide such valuable comments. It took us a lot of time in order to substantially restructure the paper and provide answers to the comments

Based on the comment by reviewer 1 ‘You are trying to squeeze two contributions into one paper: introducing a novel sense making framework (and evaluating its efficacy), as well as introducing "process-based accountability". My default recommendation is to make it two papers so that each can very clearly focus on the one point it is making. It is then much easier for you to have a targeted introduction and literature review’.

So, we follow this critical suggestion, and we focus on this novel decision making framework. At the same time, we avoid accountability and the additional layer of complications that such analysis can cause. This reduces also the length of the paper while allowing us have a more targeted introduction and literature review.

This also gave us the opportunity to narrow the focus to the Knightian distinction of risk and uncertainty in relation to knowledge and expertise and provide also a raised and specific framework that can be used for decision making.

We answer to every suggestion that the reviewers had and we hope to have done in an effective manner, in a much changed paper.

This is still a paper with weaknesses, due to the massie restricting but we hope to receive further useful comments to improve it.

REVIEWER 1

Thank you for an interesting paper. I am offering a number of suggestions for your consideration in the following. I very much like the intend behind your paper, but I believe it needs significant additional work.

Abstract / General Remarks:

First of all, thank you for the time and effort to read our paper and provide such valuable comments.

We base our discussion and analysis mainly on the work and definition of Knight (1921) ‘To preserve the distinction which has been drawn in the last chapter between the measurable uncertainty and an unmeasurable one we may use the term ‘’risk’’ to designate the former and the term "uncertainty" for the latter.’ (Knight, 1921), to avoid confusion with other studies. In this revision we make that explicitly in a variety of parts.

We added this definition and a more extensive explanation and analysis of Knightian distinction of uncertainty and risk in lines

Knight mainly focuses on probabilities so we refer to probabilities in our paper (rather than specific event and its impact that is more indirectly implied).

We also add a paragraph on knowledge relating knowledge with uncertainty.

Thank you We have simplified that, (see above) to focus only on probabilities.

Thank you, we focused on the broader term of ‘managing risks’.

- You are trying to squeeze two contributions into one paper: introducing a novel sense making framework (and evaluating its efficacy), as well as introducing "process-based accountability". My default recommendation is to make it two papers so that each can very clearly focus on the one point it is making. It is then much easier for you to have a targeted introduction and literature review. It is OK to publish two papers out of one research study.

Thank you. This was by far the most useful comment. Repeating General comments provided above:

We would like to thank both the editors and in particular the reviews for their time and substantial effort to provide such valuable comments. It took us some time in order because of the major revisions and to restructure the paper.

So, we follow this critical suggestion, and we focus on this novel sense making framework. At the same time, we avoid accountability and the additional layer of complications that such analysis can cause. This reduces also the length of the paper while allowing us have a more targeted introduction and literature review.

- Abstract, line 23, "The study also highlights...": You are making the same points three times in a row - line 23-30 should be two lines

We have in general cut, rephrased and refocused the abstract to almost half its original size and avoid repetition.

Introduction:

- Line 33 - 105: I did not count how many assertions you make, I would estimate around 25-30. That requires 25-30 literature references. I counted the current literature references: 0 .

Thank you, we mainly cut the assertions, simplified the paper and added some references.

Thank you. This is a main omission and waters down the arguments by not properly justifying them. We added significant number of references to substantiate our opinion and partly rewriting these lines to capture your suggestions.

- Line 72: Here you start the "second story" that I believe belongs into a second paper

Thank you – we cut down on accountability (belonging rather to a second paper) and refocused on the framework (see also eneral comments).

Literature Review:

Thank you for these comments. Now that we have cut the paper we can go more in greater depth.

Firstly, we present in more detail the concepts and distinction by Knight (1921).

This is a rather wide topic and we try to capture this broadly rather more in depth. However, we certainly recognize the validity of this comment. Therefore we have partly restructure and rewritten the whole literature review adding more relevant literature and most importantly analyse it in a more critical manner.

- 2.7, your critique of ISO31000. This has also been discussed in the literature. The section is mostly a summary of your reasoned opinions, not a literature review.

We revised and restructured this section.

- 2.9 / 2.10 - as I said before, this is a second paper.

Thank you once again – we cut down 2.9 and 2.10.

Methodology:

Thank you, we cut all these down and focused on risk matrices.

- 3.2: The origins of the rubrics is not clear to me. Why these? Why are they complete?

We cut the rubrics and we focus on other methods, i.e. matrices.

We omit the case study and rather suggest a theoretical model.

- 3.4: How is the refinement done? I assume this also relates to the previous theory development? What theory building framework are you building on?

We revised and restructured this section.

Results and Framework:

Thank you. We focus on Matrix methodology, please see part on Matrices on Risk and Knowledge

- 4.2: Quadrant description: You are describing the content of quadrants based on... what? Has any literature been mapped into the quadrants? Are you hypothesising what may be in the quadrants?

Thank you this is a remarkably useful comment that helped us a lot. We use related literature and existing studies on matrices to map the quadrants, of course changing and innovating in relation to previous studies.

We cut entirely rubric development.

- 4.4: These are "illustrative examples without empirical data or references" at best. These are NOT case studies.

Our papers changed focus and it became theoretical.

Thank you, we did so by focusing on a theoretical framework

Accountability is cut and 4.5 is generally changed.

- The two sections 5 (conclusions and study novelty...) need to be thoroughly revised based on an updated manuscript.

Thank you we did so.

Reviewer 2 Report

Comments and Suggestions for Authors

See attached review.

Comments for author File: Comments.pdf

Author Response

First, I would like to thank the authors and the editor for the opportunity to review this interesting and thought-provoking paper. It deserves to be published in some form. I am recommending major revision rather than minor revision not because the paper is seriously flawed, or because I expect the authors to adopt all of my suggestions (for example, even though I disagree with the Knightian dichotomy, it's certainly fine for papers to be published using that framework), but because I would like to push the authors to think more deeply about the ideas they are promoting. I should also note that I come from a different background than the authors, approaching this paper as a decision theorist, which may explain some of my different views, but also hopefully helps me approach the paper in an orthogonal way, seeing different strengths and weaknesses than the authors do. I also want to say that the first half of the paper is much stronger than the second half, so perhaps a shorter version of the paper would present the authors’ views in a better light and be more impactful.

Thank you for the useful comments and suggestion. We very much appreciate them.

Based also on the other reviewer’s suggestion this paper makes two distinct contributions that can be developed to two papers. So, we significantly reduce the size of the paper and focus on uncertainty and knowledge and cut accountability and other concepts. See initial general comments.

To begin, as I said, I am not a fan of the Knightian distinction between risk and uncertainty. To my mind, these are not qualitative differences of kind, but rather reflect a gradation, or a matter of degrees. There are few important problems where the relevant probabilities are truly “known,” and even in cases that appear to be well handled by the frequentist theory of probability, events that appear to be “replaceable” or equivalent may not truly be so. Moreover, experience with the subjectivist theory of probability (and efforts to quantify the accuracy of subjective judgments, such as the work of Cooke and colleagues — e.g., [link] — and [link]) has shown that subjective judgment can often provide informative quantitative estimates of supposedly “unknown” probabilities even when direct statistical evidence is unavailable. Therefore, I take issue with the first sentence of the abstract, and would claim that uncertainty can often (but not always) be managed through “traditional risk management frameworks.” Likewise, the description of uncertainty as “inherently unpredictable” at the top of page 4 is too strong.

Thank you we corrected the sentence “inherently unpredictable” to ‘can be largely unpredictable’. We appreciate this comment about Cooke and related work. We in a sense believe the same, that knowledge and expertise in many cases can lead to understand the distinction of risk and uncertainty but also make some ‘reasonable’ estimations when these probabilities are unknown. However this depends on the specific event and circumstances. If you think we should expand on that in our new version we would be more than happy to do so.

To my mind, the big challenge is not known versus unknown probabilities, but rather whether the range of possible outcomes can even be specified with any clarity. When the possible outcomes can be clearly summarised (e.g., gains or losses on a financial portfolio), subjective probability estimation can provide a good basis for decision-making. However, some problems (e.g., climate change, species extinction, risks of artificial intelligence) may be so open-ended that experts cannot even enumerate or describe all possible outcomes. In this context, subjectivist probability theory is not sufficient to make the problems reasonably tractable. I would really encourage the authors to think deeply about the issue of unknown probabilities vs. unknown outcomes, and reevaluate whether Knightian uncertainty is the main concern.

Exactly, we definitely agree with that, actually the blurring refers to ‘whether the range of possible outcomes can even be specified with any clarity’.

We avoid outcomes and focus on probabilities, this is based and expanded on the section about Knight. We understand that this approach might not be enough, but we have to make some assumptions based on Knight’s work.

Another key point of the paper (the need to distinguish between good decisions and good outcomes) is an important one, but it is already well understood in decision analysis. See, for example, [link]. This idea is not novel (and therefore perhaps not a “core contribution,” as discussed on page 32), even if it is important and under-recognised. Note, however, that the need to focus on good decisions rather than good outcomes holds even when we are in a situation of risk rather than uncertainty. For example, if the probability of a bad outcome is sufficiently low, the decision not to protect against it may be a perfectly rational and appropriate one, even if the bad outcome eventually occurs. Likewise, in protecting against, say, earthquakes or floods, engineers need to choose what severity of earthquake or flood they are protecting against, and which are “beyond the design basis.” This means that even prudently designed facilities can still experience bad outcomes if an event occurs that the designers

Thank you for this great insight. We, however, simplify our analysis and avoid outcomes but rather focus on probabilities.

Consciously chose not to accommodate. (This point is also relevant to the first sentence in Section 2.9.2. I am not sure that “accountability for risks with known impact and likelihood can be clearly defined” without understanding the difference between good decisions and good outcomes.)

We cut accountability and related concepts and sections.

I also think it’s important to distinguish between substantive expertise (e.g., knowledge about COVID or the financial markets) and normative expertise (knowledge about how to express uncertainties probabilistically). See [link]. The authors’ gripes are with the limitations of substantive expertise. Making this clear early in the paper would help to advance the authors’ arguments. This again becomes relevant in the paragraph on COVID on page 19. Decision makers generally had access to substantive expertise from epidemiologists and economists, but not to normative expertise (e.g., about dealing with uncertainties, or trading off lives lost and economic damage): [link].

Thank you this is a really useful comment. We present substantive expertise early, in the introduction.

Detailed Technical Comments

On page 3, the authors state that EUT “left much to be explored regarding decisions made under uncertainty.” The fact that SEU is then not introduced until page 5 leaves readers with an overly negative impression. Few decision theorists today would propose to use EUT without SEU. The two would be better presented as a sequence of historical developments, rather than two separate methods, with one having significant weaknesses.

We expanded on subjective probabilities. However we believe that it might be a slippery slope if we further expand on it. This is because with subjective probabilities estimations of uncertain events can be, to an extent justified.

The long paragraph on page 5 is disorganised, since it jumps between theoretical and behavioural approaches. EUT (line 23) and SEU (line 223) are designed to be axiomatic and prescriptive methods for how people SHOULD make decisions. By contrast, prospect theory (line 218) and ambiguity aversion (line 230) are behavioural theories that are intended to describe how people DO make decisions. These two approaches should really be discussed in two separate paragraphs, not interleaved in a single paragraph that jumps back and forth between the two.

We have restructured the paper, but we would be happy to further amend it if you recommend it.

The discussion of large-scale projects is similarly disorganised, with Petroski introduced on page 7, but Flyvbjerg not until page 10. This becomes repetitive, coming up again on page 13 and then on page 14. The discussions of climate science and AI on page 13 are repetitive as well. Perhaps these kinds of points could be made in one place, not divided up between multiple subsections of the paper.

Thank you, we have restructured these sections to be more consistent.

One point that seriously weakens the authors’ credibility is the continued reliance on value at risk as a standard method of decision-making. It is known that value at risk is not a “coherent” measure—i.e., not consistent with axiomatic decision theory. See, for example, [link]. This would not be so problematic if VAR was mentioned only once, as an aside, but coming back to it continually ended up making me more and more exasperated, and will undoubtedly do the same with other experts on state-of-the-art financial risk management. Likewise, Jorion (2006) is not a good reference on page 18 for the same reason.
A similar challenge is the discussion of groupthink on page 7. There are known methods, such as Delphi and the work of Roger Cooke and colleagues, that address the issue of groupthink, either by reaching a consensus judgement in a mathematical manner without interaction among the...Experts or by structuring the interactions in a way that minimises groupthink. Like VAR, this is somewhat of a straw-man argument — discussing problems that have known solutions, but not crediting the solutions.

We use VAR because one of the reasons for the 2008 crisis was its failure as wel las the failure of related models. There is an extensie literature on that, that we prefer not to include and not expand on further issues as Delphi and Roger Cooke’s work. However in the new draft we are open for such addition, if you think they still fit.

On page 8, I would recommend expanding the discussion of Taleb to include the work of Kahneman on stock brokers — e.g., [link].
Another point you may want to add on page 8, in the discussion of ecology and environment, is the case of blue-ear pig disease in China:

[link]
[link].
In this case, experts were aware that the existence of a pig disease caused a risk of fraud or contamination in heparin. However, the experts had a wrong prediction of the nature of the fraud. (Experts had been expecting the use of another animal rather than pigs in the production of heparin, and had been testing for animal proteins, rather than for chemical adulteration.) This is another case where the complex adaptive nature of the supply chain caused unexpected outcomes.

With regard to the discussion of supply-chain disruptions on page 11, you may wish to discuss the work of David Woods on resilience and adaptive approaches. See, for example:

[link]
[link]
[link].
In the discussion of adaptation on page 11 (and the discussion of accountability on page 31), you may also wish to cite [link] with regard to adaptability.

We have cut down on both accountability and adaptability.

Relevant to the discussion of environment on page 14, Woods (cited above) notes that natural systems (e.g., the design of biological organisms) usually involve extensive redundancy and resiliency, since any organism not designed to be resilient would have been likely to die out rather than survive during the process of evolution.
On page 11, I was surprised at the discussion of risk management beginning on line 546. This seems extremely limited. It excludes ideas like purchasing insurance and/or revising the actual system based on the results of risk analysis, to reduce or eliminate key vulnerabilities.

Thank you. We have extended the discussion of risk management especially concerning matrices. We would be glad to include also risk management solution, as insurance etc, in the next round if you suggest so.

At the bottom of page 11, I am not sure I agree that the Columbia problem was one of unknown unknowns. Seems to me that there were engineers in the Columbia launch decision process...

Who were well aware that launching at cold temperatures might pose a risk, and were just unable to persuade management to act on that information.

On page 24, I am not sure that I understand even the basic concept of what ignorance means in context. Dunning-Kruger is usually used to refer to people with extremely limited knowledge. Overconfidence is a problem among experts, but I would hardly characterise that as ignorance.
Likewise, on page 15, what is the difference between “ignorance... about the detailed effects of climate change” vs. “lack of knowledge about the detailed effects of climate change”? You seem to be defining ignorance as having some special properties, beyond just expert lack of knowledge in fields that are not yet fully mature. I am not sure I accept that claim. In any case, it has not been made convincingly, which causes a problem later on page 21. (Further down on page 15, I wonder whether Zhu et al. is the wrong reference on line 761. I didn’t check the paper, but the title appears to have nothing to do with climate change.)

Thank you, we have thought carefully about your argumenst and we rather tend to avoid the term ignorance, and cut down on it, since it can further complicate the analysis and not facilitate a generic matrix hat we propose.

The discussion of accountability in Section 2.9 seems to be missing the entire concept of political accountability for government decision-makers, even though this is directly relevant to the last paragraph of the section. Also, I would note that there is generally a trade-off between expertise and accountability. For example, physicians are typically allowed a great deal of discretion (e.g., using drugs off-label, trying experimental treatments or surgeries), while less expert practitioners such as physician assistants may be expected to adhere to a strict set of protocols limiting the autonomy with which they are allowed to practise.

Thank you, we cut on accountability.

The table on page 22 makes it seem like part of the difference between the upper two quadrants is the presence or absence of a mathematical model (since the left quadrant refers to “Overconfidence in models”). However, I wonder whether the discussion of models may be misleading in this context. For example, an expert geologist may have a good physical understanding of geology, without ever writing down a mathematical model of the risk to be managed, but still be overconfident about the possible impact of physical phenomena that have not been considered.

Rubrics

I am not convinced that the discussion in Section 4.3 rises to the level that would describe as “rubrics.” The distinctions seem to fall short of a “scoring guide.” Maybe this material would seem more like a rubric if it were presented in table format, rather than discursively in the text. However, this also seems like the point at which the paper became weaker, and I’m not sure the description of the model as “Rubric-Driven” on page 30 is accurate. Perhaps the paper should end at this point, with the authors developing the rubrics and case studies more thoroughly for a later paper (or papers).

Thank you, we cut on rubrics and rather focus on matrices.

Case Studies

I found the case studies to be among the weaker parts of the paper. They are too brief to even really be called “case studies,” and elide important issues and distinctions. Each one of these could easily be a paper on its own. It was also unclear how a discussion of only three case studies could adequately explicate...

A model with 4 quadrants; it might be helpful to show this graphically, to illustrate how the case studies transitioned from one quadrant to another over time.

We rather changes the case studies to examples , just to highlight some practical issues and challenges to the distinction between risk and uncertainty.

Thank you we illustrate the 4 quadrants graphically on a matrix.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

I appreciate the effort that went into rewriting the paper.

However it still suffers from the two major challenges that held back the first paper - insufficient grounding in the relevant literature (even though this improved), and an unclear / unstructured development of your main contributions (theoretically or empirically - in this case theoretically).

I do not recommend another revision at this point.

For a mostly conceptual paper that strives to develop new theory, I would have expected, for example, to see a much broader review of the literature, not "just" a focus on selected key references (some of which are discussed over a page or so).

Author Response

Thank you for the review.

The manuscript is a conceptual study with a focused scope. The literature anchors the argument on definitions of risk versus uncertainty, decision theory, behavioural and cognitive bias, and the role of knowledge and expertise. The text already expands coverage across Knight, Keynes, Ellsberg, behavioural decision research, engineering judgement, megaproject risk, structured expert judgement, and risk governance standards. The treatment stays selective by design to keep the theory tight.

The contribution is already explicit. The paper defines the problem as misclassification of uncertainty as risk, then offers a quadrant framework linking knowledge, expertise, and the risk–uncertainty divide. The grid is presented as a classification tool for decision contexts, not a frequency–severity risk matrix. Table 1 is labelled and discussed accordingly. The discussion already introduces decision checkpoints and cross-disciplinary consultation.

Clarity and balance have been improved within the current scope. Section 2.1 uses direct quotations and states the focus on probabilities rather than outcome ranges or uniqueness of instances. The analysis separates substantive and normative expertise, links overconfidence and calibration, and recognises limits of expert judgement. The finance section now acknowledges VaR’s incoherence and introduces CVaR for tail losses.

In short, the revision strengthens grounding and structure while preserving a theory-driven focus. Further broadening would dilute the central claim.

Reviewer 2 Report

Comments and Suggestions for Authors

I offer these comments to the authors to help them clarify their thinking, but do not expect a response to these comments, nor do I expect to review their next revision. I recognize that the authors are working from a different paradigm than mine, so I don't anticipate approving of all their editorial changes! Anyway, see below for comments.

1) As I explained in my previous review, I don't see risk and uncertainty as distinct, but as inherently on a continum (i.e., blurry). So, the phraseology that knowledge and expertise "blur the line between risk and uncertainty" (in the title and elsewhere--e.g., line 57) seems strange to me. I would rather see something like "knowledge and expertise can lead to erroneously treating uncertainty as risk."

Relatedly, on line 480, I don't see how "Specialists reduce uncertainty..." If some things are inherently unknowable in Knight's sense, I don't see how specialists can convert those uncertainties into risks. Likewise, "the blurring of risk and uncertainty" (lines 729-730) could better be described as "the confusion of risk and uncertainty."

2) Section 2.1 seems disjointed and disorganized to me. I have not gone back to check the original version, but suspect that some of the discontinuities may be due to editorial changes made since the previous draft.

For example, lines 92-93 clearly refer back to the quote on lines 87-90, but the purpose of this sentence is unclear to me. WHAT might not be known? Can the sentence be either deleted or clarified?

On lines 98-99, how is it possible to "reaffirm" a "confusion"? Perhaps this should say "continues the confusion," or similar. "Reaffirms the confusion" makes it sound as if Knight actually WANTS to be confusing.

"Therefore" on line 104 comes immediately after your statement that you focus on probabilities -- but that canot possibly be the cause of Knight's "crucial distinction." Can you please clarify exactly WHY Knight made this distinction, and what in your previous discussion explains his reasoning?

3) On line 148, I would say that EUT "assumes known probabilities," rather than "relies on..." Yes, it is an assumption, but EUT can still be useful even if probabilities are estimated rather than known. Likewise, on line 149, I would say "cannot be calculated or estimated."

4) On line 179, I would not say just "involving ambiguity," not "involving ambiguity or uncertainty." "Uncertainty" is used in many different ways throughout the paper, whereas "ambiguity" refers to only a single type of uncertainty. If you want to explain what ambiguity means in context, you could say "ambiguity (i.e., uncertainty about probabilities)."

5) Not sure why Petroski (mentioned on line 255) is not looped in again on page 11, when Flyvbjerg is discussed. Isn't the budgeting problem mentioned by Flyvbjerg just a subset of the complex engineering problems discussed by Petroski?

6) In lines 281-289, one could mention that there are methods for overcoming some of these biases (e.g., the work of Cooke mentioned in my earlier review). However, I don't feel strongly about this, since groupthing and herding are in face examples of cognitive biases.

7) With regard to the discussion of Taleb on page 7, I would argue that probabilistic models such as EUT can in fact account for extreme events or "black swans." However, I agree that these factors often are overlooked. His criticism goes too far, in my view, but has an element of truth to it.

8) I think the entire discussion of risk matrices in Section 3.2 is unnecessary and unhelpful. If you want to keep it, you should at least shorten it. Mainly, the matrix in Table 1 is not a risk matrix as commonly understood, with frequency on one axis and severity on the other -- so, the discussion of risk matrices and their popularity does not really support your use of a matrix form in Table 1.

9) At a minimum, if you keep the discussion of risk matrices, I would suggest that you delete the references to Cox (which by the way does not appear in the reference list). In particular, the main point of Cox (2008) is that risk matrices are inherently misleading. Thus, it is a misrepresentation of Cox's views to cite him in support of the idea that risk matrices are "popular" (line 369) or "well understood" (lines 402-403). In fact, I think Cox would argue that most stakeholders do NOT understand the weaknesses of risk matrices.

10) I continue to think that you need to provide a clearer distinction between normative vs. substantive expertise (e.g., https://www.mdpi.com/2079-8954/12/5/180). Many of your critiques rest on the fact that people with extensive substantive expertise may not be strong on normative expertise (i.e., knowing how much they know) -- related to metacognition. This comes up in Sections 4.1 and 4.3, and then again on page 15. I think introducing these terms would help to clarify some of your points.

11) I continue to have a concern with the paper upholding Value at Risk as a modern risk-management tool (e.g., pages 9 and 13). As noted, for example, at https://en.wikipedia.org/wiki/Value_at_risk#VaR,_CVaR,_RVaR_and_EVaR, it has long been understood to be "incoherent" (i.e., inherently irrational). I agree that it is commonly used, but is not a good tool for managing risks. If you want to cite a more modern tool, you could refer to conditional value at risk. Noting that value at risk is inadequate is sort of a straw-man argument.

I would especially suggest removing value at risk from Table 1. Since it is inherently an incoherent risk measure, it is not just "novice users" who can run into trouble using this method, but ANY user. Relying on value at risk is almost inherently an admission that the user is not truly an expert.

12) On line 457, you may want to cite "calculable or estimatable risks."

13) On lines 512-513, you say that "physicians are more likely to make errors" in complex cases. By "errors," do you mean poor decisions (i.e., decisions that someone else would have known in advance to be mistakes), or just decisions with poor outcomes (i.e., cases where the best decision was truly unknown, and a physician choice in the face of such uncertainty led to a poor outcome. If you mean the second, I would not refer to such situations as "errors."

14) On page 14, when Table 1 and line 671 recommend "consultation," consultation WITH WHOM? If the world is in a state of ignorance, who would be available as a source of consultation? Later, you introduce the idea of "cross-disciplinary consultation" (e.g., line 701), but this idea is lacking on the previous page. Likewise, there is not a clear explanation of how cross-disciplinary consultation could compensate for ignorance or "avoid catastrophic decisions." (To be clear, I think such a case could be made, but is not clearly explained currently.)

15) Early in the paper, you clarify that you are focusing on unknown probabilities, not on risks with an unknown range of possible outcomes. However, this caveat is not reiterated in Section 6.2 on study limitations, or in Section 6.3 on future research.

16) The references do not seem to have been updated. For example, Vaughan is listed in the references, but no longer discussed in the text. Conversely, Cox is discussed in text, but not listed in the references.

Minor points:

1) On line 214, "prospect theory" should not be capitalized. Same for "modern portfolio theory" and "capital asset pricing model" later in the paragraph, and "value at risk" on line 442.

2) On line 218, the comma after "Model" should be deleted.

3) Line 219 should say "provide," not "provides," since the subject of the sentence is plural (referring to both MPT and CAPM).

4) In line 366, "Matrices" is misspelled.

5) In line 469, I would say "in other fields," rather than "other fields."

6) In line 484, I think you can just say "elusive," instead of "rather elusive."

Author Response

Thanks for the extremely detailed and relevant feedback. We list below the changes made:

Wording and title
Replaced “blur the line” with “misclassify uncertainty as risk”. Title and abstract reflect this change. Main text adopts “treat uncertainty as risk” and “confusion” language. Specialists no longer “reduce uncertainty” in the Knightian sense. They reduce risk and estimation error only.
Section 2.1 clarity
Section 2.1 now quotes Knight, then interprets, then states scope. Focus on probabilities is explicit. References to outcomes and instances are framed as context, not the core distinction.
EUT phrasing
Text now reads “EUT assumes known probabilities” and notes cases where likelihoods cannot be calculated or estimated.
Ambiguity usage
Ambiguity is defined as uncertainty about probabilities. Usage aligns with this definition.
Petroski and Flyvbjerg link
The engineering section connects Petroski’s judgement failures with Flyvbjerg’s megaproject overruns.
Bias mitigation
Structured expert judgement is acknowledged. Cooke’s model is cited with calibration and performance-based weights.
Taleb and extremes
The critique now distinguishes practice from theory. The text notes fat tails, EVT, stress testing, and model error, with the main gap in specification and governance.
Matrices section
The grid is framed as a classification aid for decision contexts. It is not a frequency–severity risk matrix. Table 1 is labelled as a classification of risk and uncertainty by knowledge and likelihood.
Substantive vs normative expertise
The paper states a primary focus on substantive expertise and adds normative elements. It references Asare and Wright. It highlights calibration, base-rate anchoring, and independent challenge.
VaR and CVaR
Text now flags VaR as widely used yet incoherent. CVaR is presented as preferred for tail losses.
“Errors” in medicine
Wording shifts to poor risk assessments and suboptimal plans in complex cases. It avoids labelling outcome-driven failures as errors where knowledge is limited.
Consultation specificity
The discussion specifies cross-disciplinary consultation, normative checks, calibration, and independent challenge, with links to the ignorance-led uncertainty quadrant.

Article Menu

The Illusion of Control: How Knowledge and Expertise Misclassify Uncertainty as Risk

Rubrics

Case Studies

Further Information

Guidelines

MDPI Initiatives

Follow MDPI