The Artificial Intelligence Explanatory Trade-Off on the Logic of Discovery in Chemistry
Round 1
Reviewer 1 Report
The paper is well developed and addresses very important and original issues in the specific field of chemistry discovery improved by originally developed ML algorithms in AI systems.
However just for this relevance of the work, I would suggest the Author(s) to improve the formal component of the contribution. In the sense of inserting a graphic illustration or better a schematic formula of the type inserted in lines 378-384 summarizing each conceptual step of the overall reasoning. Otherwise, the reader at least risks "to get lost in the too many words".
Finally, from the paper it is not evident if the Author(s) developed effectively a ML algorithm of the type described in the paper and enriched with the "causal mechanism" described in the paper. If so, it would be nice to insert at least the flow chart of the routine described, if not (of course, if copyright problems do not occur) the same routine.
Finally, if the ML algorithm would be effectively developed and applied to some of the examples listed in the paper (e.g., the catalyst problem) it could be significant to discuss briefly the results of the test.
Of course, also without these modifications and improvements the paper remains intriguing and interesting, but - I repeat - too much "old fashion" philosophical, while more attention to the formal philosophy tools more recently developed could make the contribution more effective, clearer, and scientifically relevant.
Author Response
Thank you for your pertinent comments and suggestions.
Regarding the reviewer's concerns on document overall readability and English editing, I have substantially reformulated the manuscript. All changes made are highlighted in yellow in the new version of the manuscript.
"I would suggest the Author(s) to improve the formal component of the contribution. In the sense of inserting a graphic illustration or better a schematic formula of the type inserted in lines 378-384 summarizing each conceptual step of the overall reasoning."
The section that the reviewer referrers to, section 4, particularly lines 360-383, were substantially reformulated. It was added additional information on how to distinguish the QSAR from the iQSAR reasoning, and how it implicates in the schematic formula description. I believe these changes address the reviewer's concerns of overall reasoning of the presented arguments. Since I describe the meaning of each elements in equations 1 and 2 (lines 363, 367), namely chemical activity and structural property, I believe that there is sufficient information that does not require another schematic image for it.
"Finally, from the paper it is not evident if the Author(s) developed effectively a ML algorithm of the type described in the paper and enriched with the "causal mechanism" described in the paper. If so, it would be nice to insert at least the flow chart of the routine described, if not (of course, if copyright problems do not occur) the same routine"
The manuscript uses an example of other computational chemists (references 39 and 40) and is not an original model of the authors. But its value for the structured argument for a logic of discovery in chemistry is made in sections 4, 5 and 6.
"Of course, also without these modifications and improvements the paper remains intriguing and interesting, but - I repeat - too much "old fashion" philosophical, while more attention to the formal philosophy tools more recently developed could make the contribution more effective, clearer, and scientifically relevant"
Although we share the reviewer's concert on the lack of recent philosophical tools, the fact is that there is not much conceptual development regarding explanation in chemistry towards the type of problems discussed in the manuscript. However, we do have accounted the recents works of James Woodward and Lauren Ross (references 28 and 52) one the issue of causation in the exact sciences which constitute a relevant recent tool used by philosophers of science. Beyond this example, our focus was to recover older stances on the notion of induction and deduction (such as Hempel and Carmichael) as to reveal that these conceptual structures may be of use when trying to explain complex 21st century issues that arise with AI methodologies in chemistry. And since, to the best of our knowledge, this is the first work that deals with this issue on a conceptual level, we deem this manuscript as a first endeavour in coining these types of reflections.
Thank you for your time and consideration.
Reviewer 2 Report
Comments for author File: 
Comments.pdf
Author Response
Thank you for your pertinent comments and suggestions.
Regarding concerns on document overall readability and English editing, I have further copy-edited the manuscript. All changes made are highlighted in yellow in the new version of the text.
I have followed the reviewer's suggestion of making consistent changes in word formality on 'we vs I', which were corrected in the new version of the manuscript.
Also, as suggested by the reviewer, we have changed the term 'logics of discovery' to 'logic of discovery' to ensure better overall terminology. This was done throughout the manuscript and also in its title and keywords.
Thank you for your time and consideration.
Reviewer 3 Report
This is a difficult paper to review because sometimes the statements are just unclear due to linguistic issues. Several times I found statements ambiguous or obscure. To be as helpful as I conveniently can, I have uploaded the paper with my comments. I don't feel able to arrive at a verdict at this time. I have highlighted surprising word choices as well as statements that struck me as important, so the meaning of highlighting will not always be clear.
Comments for author File: Comments.pdf
Author Response
Thank you for your pertinent comments and suggestions.
"This is a difficult paper to review because sometimes the statements are just unclear due to linguistic issues. Several times I found statements ambiguous or obscure. To be as helpful as I conveniently can, I have uploaded the paper with my comments".
Regarding the reviewer's concerns on document overall readability and English editing (highlighted in the reviewer's attached file) I have substantially reformulated and copy-edited the manuscript. All changes made are highlighted in yellow in the new version of the manuscript.
"I don't feel able to arrive at a verdict at this time. I have highlighted surprising word choices as well as statements that struck me as important, so the meaning of highlighting will not always be clear"
Conceptual suggestions made by the reviewer in his commented manuscript were addressed in the new version of the manuscript. Particularly, sections that were highlighted as confusing, unclear or lack substantial arguments were reformulated (in particular sections 2, 3 and 4). I believe that the substantial changes made now adequately positions the manuscript for a positive evaluation assessment.
