# Theoretical Understanding: Bengson et al. vs Dellsén Two contemporary accounts of theoretical/objectual understanding that take different approaches to explicating what it means to understand a phenomenon. Both reject the reduction of understanding to knowledge or justified belief, but differ significantly in their structure and requirements. ## Bengson, Cuneo & Shafer-Landau: Six Properties Account From *Philosophical Methodology: From Data to Theory* (2024). ### Core Definition > "Theoretical understanding, as we'll construe it, is the state that agents possess just when they fully grasp a theory with the following six properties." (p. 28) Understanding is achieved when inquirers "fully grasp" a theory possessing specific characteristics. The focus is on *theory-possession*—what kind of theory yields understanding when grasped. ### The Six Properties #### 1. Accuracy > "First, the theory possesses a high degree of accuracy, since largely inaccurate theories will fail to dispel confusion (a characteristic of misunderstanding)." (p. 28) #### 2. Reason-Based > "Second, the theory is reason-based, in the sense that it is positively supported by considerations, beyond mere coherence, that speak in favor of its accuracy. For in the absence of such support, signing on to the theory would be arbitrary or haphazard (again, a characteristic of misunderstanding)." (pp. 28-29) This is a crucial requirement: mere internal consistency is insufficient. The theory must be *supported by reasons* that go beyond coherence. #### 3. Robust > "Third, the theory is robust, answering a multitude of questions about the most important features of the domain under investigation. A theory that neglects or dodges such questions leaves out just what's needed to yield comprehension." (p. 29) #### 4. Illuminating > "Fourth, the theory is illuminating, in that its answers must at least sometimes be not just general but also genuinely explanatory, going beyond a mere description of those features to explain why each exists or is instantiated." (p. 29) Understanding requires *genuine explanation*, not mere description. This distinguishes understanding from comprehensive description. #### 5. Orderly > "Fifth, the theory is orderly, not simply offering such feature-specific explanations but also affording a broader view of the domain by revealing how those (and other) features, as well as the proposed explanations, gel or hang together—for example, by exposing basic relations or systematic connections among them. Such a theory avoids miscellany, the paradigm of which is a mere list, which says nothing about how, if at all, its various items are ordered or organized." (p. 29) #### 6. Coherent > "Sixth, the theory is coherent, not only internally but also externally, fitting well with a wide range of understanding-providing theories of other domains. A theory of one subject matter that massively conflicts with a coherent cluster of accurate, reason-based, robust, illuminating, and orderly theories of other domains does not further comprehension but muddles it (yet another characteristic of misunderstanding)." (p. 29) ### Hierarchical Structure The six properties are not equal: > "Although all six properties contribute to theoretical understanding, they do so in different ways. The latter two, unlike the former four, only conditionally make such contributions. The orderliness and coherence of a theory contribute to its ability to supply understanding only if the theory possesses the other four features to at least some extent. In this way, these first four are fundamental to understanding in a way that the final pair are not." (p. 29) **Implication**: A theory can be orderly and coherent yet provide *no understanding* if it lacks accuracy, reason-based support, robustness, or illumination. ### Not Reducible to Knowledge > "Theoretical understanding [...] isn't reducible to true or justified belief (or to their conjunction). Nor is it equivalent to ordinary knowledge. For, as explained above, the function of such understanding is, inter alia, to illuminate, in a robust, orderly, coherent fashion, the portion of reality under investigation, and not simply to state a series of known truths or justified beliefs about it." (pp. 29-30) ### Critique of Reflective Equilibrium Bengson et al. use Reflective Equilibrium as their primary example of a method that produces orderly, coherent theories that fail to provide understanding: > "Successfully implementing the Method of Reflective Equilibrium guarantees that a theory will at least be robust, orderly, and coherent. But the method infamously fails to put inquirers on track to achieve even a modicum of accuracy; nor is it clear that it plumps for outputs that are reason-based in the indicated sense." (p. 102) The problem: RE has no mechanism for checking accuracy. If your initial judgments are wrong, RE will systematize errors into a coherent but false theory. > "If the modifications, additions, or abandonments that help inquirers achieve equilibrium are wide of the mark, lack support, or undermine robustness, then the equilibrium attained will fail to promote theoretical understanding." (p. 103) They also invoke Sellars' critique: > "The method thus opens the door to systematic theories whose claims are unsupported by any consideration, beyond coherence, that speaks in their favor. This raises the worry that the Method of Reflective Equilibrium is not adequately support-requiring. It sanctions outputs that belong 'in a box with rumors and hoaxes,' to quote Wilfrid Sellars' barb targeting a narrow focus on coherence." (p. 100) --- ## Dellsén: Dependency Modelling Account (DMA) From "Beyond Explanation: Understanding as Dependency Modelling" (*BJPS* 2020). ### Core Definition > "DMA: S understands a phenomenon, P, if and only if S grasps a sufficiently accurate and comprehensive dependency model of P (or its contextually relevant parts); S's degree of understanding of P is proportional to the accuracy and comprehensiveness of that dependency model of P (or its contextually relevant parts)." (p. 1268) ### What is a Dependency Model? A dependency model represents how phenomena stand in *dependence relations* to one another: > "The aspects of a phenomenon that matter for understanding are the dependence relations that the phenomenon, or its features, stands in towards other things. The most well-known kind of dependence relation is causality—an effect depends on its cause—but there are arguably other kinds of dependence relations as well. One of these arguably non-causal dependence relations is grounding—that is, the in-virtue-of relation." (p. 1266) Dependency models include both *positive* and *negative* information: > "Causal graphs [...] are meant to contain both 'positive' information about the causal relations that are present within a system, and also 'negative' information about which parts of the system are not causally related." (p. 1267) > "A pictorial representation of what I am calling a dependency model would be a graph which purports to depict how each element depends, or does not depend, on each other element—causally or otherwise." (p. 1267) ### Two Dimensions of Quality Understanding is measured by two gradable criteria: > "A dependency model better represents P to the extent that the network of dependence relations that P stands in is correctly depicted by the model. Since a dependency model can thus fail either by incorrectly representing (that is, misrepresenting) some aspect of this network, or by not representing it at all, we can identify two separate criteria here, namely, accuracy and comprehensiveness." (p. 1267) These can trade off: > "These two criteria can come into conflict: increasing comprehensiveness may require us to sacrifice accuracy, in which case we engage in a kind of idealization; conversely, to increase accuracy we may have to sacrifice our model's comprehensiveness, in which case we engage in abstraction." (p. 1267) ### Gradability Understanding is inherently gradable: > "One's degree of understanding of some phenomenon, P, is proportional to the comprehensiveness and accuracy of one's dependency model of P." (p. 1267-68) Binary uses of "understanding" are contextually determined thresholds: > "Such a binary concept of understanding can be explicated in terms of a degree of understanding that exceeds some threshold t, where the t is presumably contextually determined (for example, by the speaker's context of utterance)." (p. 1268) ### Understanding Without Explanation A key innovation: understanding can come apart from explanation. Dellsén identifies three types of cases: #### 1. Explanatory Bruteness Phenomena that have no explanation at all can still be understood: > "For any such fact, there is a phenomenon that is better understood if one's representation of the phenomenon includes a depiction of the very fact that makes it unexplainable, namely, that its occurrence or existence does not depend on anything else." Understanding an indeterministic system: > "The dependence relations in an indeterministic system can be modelled, no less than those in a deterministic system. Thus, if we grasp the appropriate kind of model, that is, a sufficiently accurate and comprehensive dependency model, then we understand it. Importantly, note that a model is more comprehensive if it includes rather than excludes the information that the electron's going right rather than left is a fifty-fifty chance event, the outcome of which does not depend on anything else in the system." #### 2. Explanatory Targetedness Understanding a phenomenon without having explanations of its specific features. #### 3. Explanatory Disconnectedness Understanding increases by discovering that things are *independent*: > "A third and final type of case in which understanding comes apart from explanation is when our understanding is increased by virtue of discovering that some object or feature does not explain another object or feature." (p. 1278) Galileo's reductio example: > "Our understanding increases because we come to realize that seemingly related factors—namely, mass and gravitational acceleration—are independent." (p. 1279) > "These types of cases are easily accommodated by DMA, since one way in which a dependency model of a phenomenon can be made more comprehensive is by specifying that two of its features—for example an object's gravitational acceleration and its mass—do not depend on each other." (pp. 1279-80) ### Relationship to Explanation Despite separating understanding from explanation, Dellsén acknowledges a connection: > "Since dependence relations are, at least normally, what undergirds explanations, there is still a strong connection between understanding and explanation on this view. So there is a kernel of truth in explanatory accounts of objectual understanding. However, whereas explanatory accounts take explanation to be a necessary component of such understanding, the current account allows for understanding to come apart from explanation." (p. 1269) --- ## Direct Comparison | Feature | Bengson et al. | Dellsén | |---------|---------------|---------| | **What you grasp** | A *theory* with 6 properties | A *dependency model* | | **Core structure** | 6 qualitative properties (4 fundamental + 2 conditional) | 2 quantitative dimensions | | **Properties/Dimensions** | Accuracy, reason-based, robust, illuminating, orderly, coherent | Accuracy, comprehensiveness | | **Hierarchy** | Yes—first four fundamental | No—dimensions trade off | | **Requires justification?** | Yes—"reason-based" is fundamental | No | | **Requires explanation?** | Yes—"illuminating" requires genuine explanation | No—understanding can come apart from explanation | | **Gradability** | Implicit | Explicit | | **Negative facts** | Not emphasized | Central—"does not depend on" matters | ### Key Points of Divergence **1. Role of Reasons/Justification** Bengson et al.: Reasons are *constitutive* of understanding. A theory must be "positively supported by considerations, beyond mere coherence, that speak in favor of its accuracy." Without this, grasping the theory is "arbitrary or haphazard." Dellsén: Justification *promotes* but doesn't *constitute* understanding. You can understand via an accurate dependency model regardless of whether you have reasons for thinking the model is accurate. **2. Role of Explanation** Bengson et al.: Explanation is required via the "illuminating" property—answers must be "genuinely explanatory." Dellsén: Explanation is not required. Understanding can increase by learning that something has *no* explanation, or by learning what something is *independent* of. **3. What Understanding Tracks** Bengson et al.: Understanding tracks the *epistemic quality* of a theory—is it supported, explanatory, systematic? Dellsén: Understanding tracks the *representational accuracy* of a dependency model—does it correctly depict how things depend on each other? ### Points of Convergence 1. Both reject reducing understanding to knowledge or justified belief 2. Both require *accuracy*—connection to how things actually are 3. Both value systematicity—seeing how things "hang together" 4. Both reject mere lists or accumulation of facts without structure --- ## Implications ### For Philosophical Method Bengson's account is more demanding: philosophical inquiry must produce reason-based, illuminating theories. Methods like Reflective Equilibrium that don't guarantee accuracy or reason-based support are inadequate. Dellsén's account is more permissive: any process that results in accurate dependency representations yields understanding, regardless of the method used. ### For AI-Generated Content **Neither account creates conceptual friction for AI-enhanced understanding.** The "reason-based" property is a property of *the theory*, not *the producer*. Whether reasons exist that support a theory is independent of whether the entity that produced it was "reasoning" or "understood" what it was doing. (Parallel: a calculator doesn't "understand" arithmetic, but 2+2=4 is mathematically supported regardless.) If an AI produces a theory that *is* supported by reasons, and a human grasps it, understanding is achieved. The causal history of production drops out as irrelevant. **Open empirical question**: Do AI systems tend to produce *mere coherence* (like RE outputs) or *genuinely reason-based* theories? This is contingent on training and capabilities, not a conceptual point about understanding. The structural parallel with RE's coherence-optimization is worth investigating—but it's an empirical question about what AI actually produces, not a conceptual barrier. ### For Progress Despite Disagreement Bengson's account may face challenges from persistent philosophical disagreement—if reason-based support requires consensus on reasons, disagreement might block understanding. Dellsén's account handles disagreement better: understanding doesn't require justified belief, so even where philosophers disagree about justification, they might still be making progress by developing more accurate dependency models. --- ## Sources - Bengson, J., Cuneo, T., & Shafer-Landau, R. (2024). *Philosophical Methodology: From Data to Theory*. Oxford University Press. - Dellsén, F. (2020). Beyond Explanation: Understanding as Dependency Modelling. *British Journal for the Philosophy of Science*, 71, 1261-1286. ## Related - [[Metaphilosophy Landscape]] - [[Notes/Reflective Equilibrium]]