"The problem of explaining shifting targets" - David Colaço (LMU Munich)

By David Colaço

Sometimes, we try to explain one thing but end up explaining something else. In other words, the explanatory target (or explanandum) might change. Often, this change is productive. Philosophical accounts like those I review here address when targets should be recharacterized based on new evidence, including evidence about the details that explain it (or explanans). After all, we want our explanations to be as good as possible. For this reason, the target should sometimes change, so that it and its explanation fit our best evidence. Though (hopefully) intuitive, this line of reasoning raises the possibility of a more problematic outcome: that we stick with an original target, though the details better explain another.

This is the topic of “The Problem of Explaining Shifting Targets” (2026, EJPS), in which I introduce the eponymous problem, PEST for short. PEST has three steps. First, an agent articulates a target that they want to explain. How they do this can vary, but I think of articulation as characterizing the target or facts about it. Given that ‘target’ or ‘explanandum’ can be ambiguous, I coopt Hempel’s distinction between explanandum-statement and explanandum-phenomenon. My use of ‘target’ fits the former, and for those of an erotetic disposition, they can be questions. These targets stand in relation to the latter. What matters is that, however articulated, a target has specificity. It has empirical consequences. It also has contrastive foils. We can treat the target as a fact contrasted by a foil, asking why this rather than that. How we parse ‘this’ from ‘that’ reflects the specificity of the target.

Second, when explaining the original target, an agent shifts to a manifestly different target. ‘Explaining’ denotes both how one formulates the explanatory details that answer how or why a target occurs and the investigations that support this formulation. When targets are manifestly different, evidence can be provided for one that is not evidence for another. This difference can be of specificity, such as when the original target is more detailed than the shifted one. It also can be of reference, such as when the targets pick out different explananda-phenomena. These differences can be detected, and an agent’s interest is sensitive to them.

Third, an agent treats the explanatory details as explaining the original target, not recognizing the manifest difference between it and the shifted target. Rather, the explanatory details are taken to be for the original target, with the agent not considering a change, whether targets are different, or whether they are equally interesting. Because the second step is satisfied, we can grant that the details explain the shifted target. However, they are presented as details that explain the original. It is this third step that makes PEST a problem.

If PEST seems abstract, a toy example helps with its bite. Imagine that your buddy tells you that they think drinking turmeric tea prevents cancer. This is the first step. Although they do not do scientific studies, they research that a chemical in turmeric has been shown to inhibit tumor invasion and growth in the lab. They explain that this is why drinking turmeric prevents cancer. Mulling over this response, you recognize that the studies say nothing about drinking. They were in vitro. Turmeric might have some effect, but it seems like their target has become importantly less specific through their explaining. This is the second step. Say that your buddy is steadfast, treating the details as explaining what they initially stated. Doing this, without consideration that their target and what they potentially can explain might be different, is the third step. While this example is simplified, my article addresses real-world cases, in which agents are non-experts (like this example), scientific experts in different fields, or even experts within a field. For each real-world case, the consequences of PEST were quite bad, wreaking havoc on the research in which they occurred.

My anecdotes suggest that readings of examples like this are worryingly diverse, highlighting the need for more investigations into our commitments to explanatory targets. Nonetheless, a common reaction is that agents in these cases are obviously engaging in bad reasoning. Bad reasoning is worth identifying, but what makes the problem an explanatory one?

What makes PEST a problem, I argue, ties to the nature of explanation. There are three reasons why this problem is liable to occur, which link to the steps of PEST. First, agents tend to be interested in original targets. This is why they want to explain them and might be reticent to give them up, especially when alternatives are less interesting to them. Second, targets are not static. Explaining can reshape them. By devising and assessing explanatory details, we gain information, which can provide new expectations about a target. These expectations can inform new empirical consequences and contras­tive foils. Yet, consequences and foils are how we articulate targets, giving explaining the ability to reshape them. Third, agents need not recognize when explaining reshapes a target. The information acquired through explaining need not undercut empirical support for the original target, so supporters of this target may not be in any obvious way acting irrationally if they steadfastly hold on to what interests them. For these reasons, PEST should not be all that surprising.

I hope that these reasons give the reader an impression of why combating PEST can be difficult. This difficulty, unfortunately, increases when we probe what it means to not recognize a manifestly different shift of targets. A good faith reading would be one in which the agent genuinely does not realize the difference. The example does not involve your buddy having ulterior motives. However, all that is needed for my account is that the agent does not acknowledge and address the shifted target.

On this reading, PEST is more sinister. An agent could shift targets on purpose while feign­ing ignorance or simply not acknowledging it. They can “bait-and-switch” targets, implicitly shifting from a controversial target to a comparatively banal but explainable one. This tactic could be used to concoct the veneer of support for a target that we should not accept. It would be effective if people were questionable judges of explanations, which I fear they often are. This sinister character is a good reason for more research on PEST: these tactics can be used to achieve unscrupulous aims. For this reason, we always should ask: what are we explaining, and is it what we are trying to explain? When we ask this ques­tion, we acknowledge that explanatory targets can shift.

This article was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), as part of Project Number 528370787.