The Problem of Differential Importability and Scientific Modeling
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe article is fine, but not very original. It is a good presentation of the sources, but the bibliography consists of books that were published some years ago and the author seems to primarily summarise the findings of these predecessors.
Point i is missing and some ideas are presented in a hurry.
With still a bit of work, it can be seriously improved.
Comments on the Quality of English LanguageThere are a few typos, infelicities of English, colloquialisms and rather unclear notions/ distinctions that are used. For instance:
Line 29 ‘One what ground’ instead of ‘On what ground'
Line 174 ‘has to reality seem … to …’ instead of ‘has to really seem … to …Line 306 – ‘whether nominalism about scientific or models’ instead of ‘whether nominalism about scientific models’
Lines 35-36 – i and ii – where is i? ii is at line 36
Line 158 – ‘the big worry’ (?)
Line 189 – ‘this manner of putting things’ (?)
Line 39 to ‘characterise’ a notion?
Lines 20-23 - pre-philosophically/ philosophically
Author Response
Note: New responses have been added in italics.
Comments 1: "The article is fine, but not very original. It is a good presentation of the sources, but the bibliography consists of books that were published some years ago and the author seems to primarily summarise the findings of these predecessors."
Response 1: Many thanks for your indication that the manuscript's contents lacked originality and somewhat dated sources. After a careful assessment of the original manuscript, I realized the plausibility of that indication, and have strived to compensate in section viii of the revised manuscript, where, in particular, an argument has been made in favor of (i) the standard realist's need to acknowledge predictive success as an exhaustive and central arbiter in comparisons of model importabilities (revised manuscript: lines 589-604), (ii) a holistic notion of predictive success that incorporates both the model's potential for prospectively intended experimental and prediction-related felicities, and the actual, reliable successes of predictions derived from it (revised manuscript: lines 522-529), and (iii) the standard realist's need to accept (i) based on a plausible conception of his own views on the role of truth and representational fidelity in comparing model importabilities (revised manuscript: lines 594-598) . All of (i)-(iii) are intended as introducing elements into the debate around model evaluation not fully addressed within extant contributions. None of the works discussed in the original, unrevised manuscript can be summarized in terms of all of (i)-(iii). (ii) is definitively a new proposal in the literature on predictive success and its import in model comparison and evaluation. (i) is also quite novel, since none of the critiques of the realist position in the more recent pragmatist turn seem to allow a standard realist conception of truth and representation to be compatible with model evaluation based exhaustively on predictive success. I have also added now a clear-cut statement of where the originality of the manuscript rests (revised manuscript: 627-639):
"The originality of this manuscript is reinforced by its distinctive handling of truth in model evaluation. Unlike existing frameworks that consider predictive success as one factor among many, this manuscript argues that truth is partially operationalized as actual empirical success, making empirical adequacy a provisional marker of truth. Additionally, it reinterprets truth-conducive virtues as elements contributing to prospective predictive success, thereby granting the realist a correspondence notion of truth based on philosophical grounds while offering a defeasible evidential basis rooted in empirical success. This nuanced approach fills a significant gap in the current literature (Parker, 2020; Bokulich, 2021; Sullivan, 2022), where empirical grounding and prospective success are often not given sufficient emphasis in realist evaluations. By integrating these elements, this manuscript offers a framework that aligns more tightly with the realist's commitment to truth, providing a holistic evaluative method that ties empirical success and truth as mutually reinforcing."
I have also engaged many recently cited works in section viii of the manuscript, including at least the following (see "References"):
12. Schindler, Samuel (2018). Theoretical Virtues in Science: Uncovering Reality Through Theory. Cambridge: Cambridge University Press.
14. Mizrahi, Moti (2020). Hypothesis Testing in Scientific Practice: An Empirical Study. International Studies in the Philosophy of Science 33 (1):1-21.
15.16. Murphy, Alice ; Currie, Adrian & Walsh, Kirsten (forthcoming). Aesthetics and Agency in Experiments. Australasian Journal of Philosophy
Comments 2: "Point i is missing and some ideas are presented in a hurry"
Response 2: Many thanks for this comment. Please accept my sincere apologies for not being able to understand what precisely is meant by "point i". However, your concern about the hurried developed in the manuscript is very pertinent, since in its original state the discussion developed rather quickly, and under a number of simplifying or questionable assumptions. I have tried my best to correct this flaw in section viii of the revised manuscript. Please refer to lines 310-350 and end-note "iii" of the revised manuscript for the details on how the assumptions are discharged over the course of the mentioned section.
Comments 3: "With still a bit of work, it can be seriously improved."
Response 3: I am very grateful for this encouraging comment, given this is my first ever attempt at publication. I hope the changes reported in Response 1 and Response 2 would be considered at least a marginal step forward towards a better piece of work than has been the case originally.
Comments 4: "There are a few typos, infelicities of English, colloquialisms and rather unclear notions/ distinctions that are used. For instance:
Line 29 ‘One what ground’ instead of ‘On what ground'
Line 174 ‘has to reality seem … to …’ instead of ‘has to really seem … to …Line 306 – ‘whether nominalism about scientific or models’ instead of ‘whether nominalism about scientific models’
Lines 35-36 – i and ii – where is i? ii is at line 36
Line 158 – ‘the big worry’ (?)
Line 189 – ‘this manner of putting things’ (?)
Line 39 to ‘characterise’ a notion?
Lines 20-23 - pre-philosophically/ philosophically"
"
Response 4: All of the above infelicities have been either corrected or removed from the revised manuscript. In particular, see lines 24, 33, 162, 180, 188, 189 for the effected corrections.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper compares three ontological views about models (abstractionism, direct and indirect fictionalism) with regards to how they account for the "importability" of models, i.e. what make them adequate for use.
The paper is clear overall. However, I think the whole discussion could be improved by analysing more tightly the link between ontological concerns (what a model is) and practical or epistemological concerns (what makes a model, or a modelling activity, suitable for use), and by distinguishing questions of ontology and questions regarding the model--object relationship. Regarding the latter, different model ontologies might favour different answers to how they related to the target of representation, but it's not so straightforward, and it is not very clear to me why one could not be at the same time an indirect fictionalist and reject a similarity view in favour of, say, a mere stipulationist account of the model--object relationship (such as that of Craig Calender and Jonathan Cohen in their "there's no problem of scientific representation) for instance.
Actually, it seems that many people today accept that what grounds the representation relation is not an actual similarity, but a posited similarity, in particular since Mauricio Suarez's criticisms of the former in his "Against Similarity and Isomorphism". This is the case, as far as I know, of Weisberg, Giere and van Fraassen. This undercuts part of the discussion from p. 4.
Independently of this, the point discussion (in particular on p.4) is not very clear to me. Why couldn't one respond that there is are measured of similarity according to various purposes (a model is more or less similar with respect to the aspects that are relevant for such or such purpose), and that the model that one should use is the most similar one available?
Regarding the threat of instrumentalism (second part of p.4), the standard realist way of viewing things is that instrumentalism is implausible in view of scientific practice, because predictions is not all that matter. Other virtues (simplicity, elegance, theoretical coherence, etc.) also matter in model construction. The paper just states the contrary without discussion. So, again, a deeper discussion of these issues would be welcome.
Finally, I'm not sure that the three views discussed exhaust the space of possibilities. Models can be either existent or not. If they exist, they can be either concrete or abstract. And if they are abstract, they can be either fictional or actual. At least one view is missing: the view that models are actual entities. This is the view of models as epistemic artefacts proposed by Tarja Knuuttila in her "imagination extended and embedded".
Author Response
Note: New responses have been added in italics.
Comments 1: "The paper compares three ontological views about models (abstractionism, direct and indirect fictionalism) with regards to how they account for the "importability" of models, i.e. what make them adequate for use.
The paper is clear overall. However, I think the whole discussion could be improved by analysing more tightly the link between ontological concerns (what a model is) and practical or epistemological concerns (what makes a model, or a modelling activity, suitable for use), and by distinguishing questions of ontology and questions regarding the model--object relationship. Regarding the latter, different model ontologies might favour different answers to how they related to the target of representation, but it's not so straightforward, and it is not very clear to me why one could not be at the same time an indirect fictionalist and reject a similarity view in favour of, say, a mere stipulationist account of the model--object relationship (such as that of Craig Calender and Jonathan Cohen in their "there's no problem of scientific representation) for instance.
Actually, it seems that many people today accept that what grounds the representation relation is not an actual similarity, but a posited similarity, in particular since Mauricio Suarez's criticisms of the former in his "Against Similarity and Isomorphism". This is the case, as far as I know, of Weisberg, Giere and van Fraassen. This undercuts part of the discussion from p. 4.
Independently of this, the point discussion (in particular on p.4) is not very clear to me. Why couldn't one respond that there is are measured of similarity according to various purposes (a model is more or less similar with respect to the aspects that are relevant for such or such purpose), and that the model that one should use is the most similar one available?
Finally, I'm not sure that the three views discussed exhaust the space of possibilities. Models can be either existent or not. If they exist, they can be either concrete or abstract. And if they are abstract, they can be either fictional or actual. At least one view is missing: the view that models are actual entities. This is the view of models as epistemic artefacts proposed by Tarja Knuuttila in her "imagination extended and embedded"."
Response 1: Many thanks for these very kind, sharp and detailed comments. I have strived to compensate for the lucanae reported by the comments, having translated them into number of simplifying or questionable assumptions that very clearly constrained the discussions within the original manuscript. Accordingly, I have tried my best to correct this flaw in section viii of the revised manuscript. Please refer to lines 310-350 and end-note "iii" of the revised manuscript for the details on how the assumptions are discharged over the course of the mentioned section. For your convenience, here is a quote of endnote "iii":
"
Discharging assumption (A1): Initially, differential importability was constrained by views that strictly conceptualized models as abstract or fictional entities. The discussion has been expanded to consider how models function in practice, regardless of their ontological status. By demonstrating that the utility of models in scientific practice often transcends their ontological categorization, this section discharges the narrow focus of Assumption (A1) and shows that practical success in modeling is often more influential than strict ontological classifications.
Discharging Assumption (A2): The assumption that differential importability requires a solution rooted solely in the ontology of models has been reevaluated. By incorporating pragmatic considerations such as the applicability, coherence, and explanatory power of models, the argument shows that practical concerns can provide a robust basis for model selection and use. This pragmatic shift discharges Assumption (A2) by proving that ontological purism is less relevant than a model's operational success in explaining and predicting phenomena.
Discharging Assumption (A3): This assumption posited that the problem of differential importability could only be resolved within the confines of existing ontological frameworks about models. The section challenges this by arguing that the epistemological and practical achievements of models often outweigh their ontological foundations. By focusing on the results and impacts of models in scientific inquiry, rather than their theoretical underpinnings, Assumption (A3) is discharged, advocating for a more flexible and results-oriented approach in scientific modeling.
Discharging Assumption (A4): Initially, there was an implied inevitability that ontological discussions about models would directly influence their philosophical appraisal in terms of representation. The developed argument shows that it is possible to discuss and evaluate models based on their epistemological roles independently of their ontological claims. This approach discharges Assumption (A4) by decoupling philosophical debates about the nature of models from practical assessments of their utility and success."
Comments 2: "Regarding the threat of instrumentalism (second part of p.4), the standard realist way of viewing things is that instrumentalism is implausible in view of scientific practice, because predictions is not all that matter. Other virtues (simplicity, elegance, theoretical coherence, etc.) also matter in model construction. The paper just states the contrary without discussion. So, again, a deeper discussion of these issues would be welcome."
Response 2: Inspired by the relevance and plausibility of this suggestion, in section viii of the revised manuscript, I have attempted to make an argument in favor of (i) the standard realist's need to acknowledge predictive success as an exhaustive and central arbiter in comparisons of model importabilities (revised manuscript: lines 589-604), (ii) a holistic notion of predictive success that incorporates both the model's potential for prospectively intended experimental and prediction-related felicities, and the actual, reliable successes of predictions derived from it (revised manuscript: lines 522-529), and (iii) the standard realist's need to accept (i) based on a plausible conception of his own views on the role of truth and representational fidelity in comparing model importabilities (revised manuscript: lines 594-598) . All of (i)-(iii) are intended as introducing elements into the debate around model evaluation not fully addressed within extant contributions.
I have carefully considered your request for clearer integration between ontological and epistemological concerns and have directly addressed this in Section viii by discharging assumptions (A1-A4). This section provides a more pragmatic and flexible framework for resolving the problem of differential importability by moving beyond narrow ontological concerns and focusing on the pragmatic performance of models.
1. Discharging Assumption (A1): Initially, I had assumed that the problem of differential importability could be reduced to the challenge of explaining differences in model importability based on their ontological status as abstract entities or entitive fictions. As noted in the manuscript, "Assumption (A1) implied that the problem of differential importability is 'just' the infelicity of accounting for differences in model importability, given some construal of models as abstract entities, or entitive fictions" (lines 335-337). I have revised this assumption to highlight how pragmatic success in predictions can diverge from representational fidelity. I wrote: "Ranking models based on their representational fidelity to the target can, in principle, disagree with their ranking based on the (reliable) success of the predictions they afford about the target" (lines 342-344). This clarification directly responds to your concern by underscoring that predictive success offers a more practical and empirically grounded measure for evaluating models than fidelity alone.
2. Discharging Assumption (A2): I had also assumed that comparative differences in fidelity and success would be tightly bound to the correspondence model-target relation. However, in revisiting this, I acknowledged that success and fidelity can diverge: "The assumption (A2), to be precise, had been the above-mentioned assumption about comparative fidelity-success divergence and the correspondence notion of model-target relations of representation" (lines 347-349). By emphasizing predictive success over representational fidelity, I have reoriented the discussion toward more pragmatic concerns, thereby better linking epistemological with ontological aspects, in line with your recommendation.
3. Discharging Assumption (A3): Initially, I overemphasized the role of ontology in resolving the problem of differential importability. In revising this, I argue that epistemological achievements, such as a model’s success in predictions, take precedence: "Assumption (A3) had been the assumption the ontology of models has some uniquely special role to play in sustaining, or resolving, the problem of differential importability" (lines 349-351). This addresses your comment by making it clear that pragmatic success in generating accurate predictions offers a stronger criterion than ontological status.
4. Discharging Assumption (A4): I have also reconsidered the assumption that the ontologies of models are inseparable from how model-target relations are construed. I now argue that these ontologies do not have to dictate the significance of the representational relations: "Assumption (A4) has been the view that the three ontologies of models under discussion are inseparable from the way in which some of their de facto proponents construe model-target relations" (lines 354-356). This shift allows for a clearer distinction between ontological commitments and practical success, which is the core of what you requested.
By discharging these assumptions, Section viii now offers a more methodologically flexible and pragmatic approach to resolving the problem of differential importability, focusing on the centrality of predictive success over representational fidelity. This directly aligns with your call for a clearer connection between ontology and epistemology.
You also asked for a more detailed engagement with instrumentalism and the role of predictive success in evaluating models. I have responded to this in Section viii by clarifying that predictive success is the exhaustive criterion for determining model importability, while also incorporating truth-conducive virtues like simplicity and coherence within a holistic framework.
1. Predictive Success as the Exhaustive Criterion: I argue that predictive success—both actual and potential—should be the ultimate measure for evaluating models. In the manuscript, I state: "The ultimate decider of how importable a model is, even if its representations are parsed in informational terms, are the successful predictions derived from the model, or the potential to generate more such predictions: with both the actual success, and the potential for success, co-constituting a holistic metric of predictive success" (lines 538-542). This aligns directly with your call for a stronger focus on predictive outcomes as the central evaluative criterion, particularly within realist frameworks.
2. Holistic Notion of Predictive Success: I also developed a holistic notion of predictive success that encompasses both actual empirical outcomes and the model’s potential for future predictive success. As written in the manuscript: "Comparatively judging model-importabilities in terms of representational fidelity in a provisional sense justified by empirical, predictive success, can be 'described' entirely in terms of notions that exclude that of 'truth', or 'fidelity'" (lines 568-570). This approach emphasizes that truth-conducive virtues like simplicity and coherence are subordinate to predictive success, and are only relevant when the predictive success of models needs further distinctions.
3. Realist’s Commitment to Truth: Finally, I explicitly tied the realist’s commitment to truth with the necessity of predictive success, stating: "The realist is compelled to conceive as much on the basis of what may plausibly be considered his own premise: that one is, in comparing model-importabilities, 'comparing' the 'degrees' of fidelity or truth possessed by the relevant models" (lines 655-657). Here, I am implying that while virtues such as simplicity may play a role, they ultimately serve a secondary function, with predictive performance remaining the key evaluative criterion, and this is ensured by the realist's own premise and the need for operationalizing objects whose "degrees" are "compared".
In addressing instrumentalism and model evaluation, I emphasize that predictive success is the key to evaluating model importability, with truth-conducive virtues playing a supportive role. This provides a comprehensive and nuanced response to your concerns.