Adrian Ziółkowski and Maciej Tarnowski
Affiliation: University of Warsaw, Jagiellonian University
Category: Philosophy
Keywords: Side-effect Effect, Epistemic Side-effect Effect, folk psychology, intentional action, intentionality attributions, knowledge attributions
Date: Tuesday 2nd of September
Time: 14:30
Location: Room 154 (154)
View the full session: Blame & the Knobe Effect
The Side-Effect Effect (SEE), or Knobe Effect (after Knobe [2003]), is a well-documented asymmetry in folk attributions of intentionality, blame/praise, and epistemic states (sometimes referred to separately as an Epistemic SEE – ESEE; Beebe & Buckwalter 2010, Beebe 2012), where negative side-effects are judged more intentional (resp. blameworthy or known) than positive ones. Studies confirm that SEE is robust across cultures, languages, and various contexts—including moral, aesthetic, and legal considerations (Cova, 2016). Since this asymmetry challenges the traditional distinction between intentionality and moral valence (Knobe, 2004), explaining its origins and normative significance remains a key task for philosophers of action, language, epistemologists, and ethicians. As research on SEE and ESEE enters its third decade, there is still no widely accepted explanation of the psychological mechanisms that trigger these effects or their broader implications. Current explanations can be roughly categorized into three approaches: Evaluative (EA), Attitudinal (AA), and Linguistic (LA). The Evaluative Approach (EA) argues that the content of the relevant concept is shaped by evaluative considerations, meaning that moral judgments influence attributions of intentionality (Knobe, 2007; Holton, 2010). The Attitudinal Approach (AA) proposes that folk judgments result from heuristics used in propositional attitude ascriptions, which may be systematically biased or faulty (Alfano et al., 2012). Lastly, the Linguistic Approach (LA) suggests that SEE arises from how linguistic intuitions shape responses to the specific wording of vignettes rather than from deeper conceptual or psychological mechanisms (Mizumoto, 2018). In our talk, we propose to take a step back and examine boundary cases where these explanatory models predict that SEE should disappear rather than persist. Specifically, we test a prediction common to EA and AA but not necessarily to LA: The Knobe asymmetry in judgments (intentionality, blame/praise, knowledge, and belief ascriptions) should not occur in cases where the protagonist cannot be meaningfully ascribed relevant attitudes (e.g., beliefs, desires, or moral agency) regarding the side effects of their actions. To assess this prediction, we conducted two studies. Study 1 examined whether SEE and ESEE persist when protagonists are explicitly described as unaware of the side-effects of their actions. If SEE is driven by evaluative considerations (EA) or by heuristics in propositional attitude ascriptions (AA), then it should diminish or disappear when protagonists lack awareness of potential side-effects. Study 2 investigated whether SEE persists in vignettes featuring non-human protagonists, such as animals other than humans, physical events, or inanimate objects, which presumably lack complex propositional attitudes. If SEE relies on attributing mental states to agents, then it should not arise in cases where no such states can reasonably be ascribed. Both studies consisted of two phases: the exploratory phase, in which we established initial findings, and the replication phase, in which we ran pre-registered experiments to corroborate the results. Our findings directly challenge existing explanatory frameworks. Contrary to the predictions of EA and AA, Study 1 found evidence of SEE and ESEE even if participants were explicitly informed that the protagonist in the scenario was unaware of the side-effects of their action. Interestingly, although this information affected respondents’ judgments in the predicted way – i.e., the overall tendency to attribute intentional states to the protagonists significantly decreased – it did not remove the asymmetry in attributions between positive and negative conditions (the result concerning intentionality judgments is illustrated in Figure 1). This suggests that considerations of belief or desire play a lesser role in explaining SEE than previously thought. In Study 2, encouraged by the abovementioned findings, we explored whether the Knobe asymmetry would emerge in judgments regarding entities (such as non-human animals or inanimate objects) that are prima facie incapable of being aware of far-reaching side-effects of the events they are involved in. Most surprisingly, this study revealed an inverse SEE and ESEE for cases where the protagonist was a non-human entity—that is, respondents were less likely to attribute intentionality or knowledge to protagonists bringing about negative side-effects than to those bringing about positive ones (the resultsfor intentionality ratings are illustrated in Figure 2). This reversal is not predicted by any of the explanatory approaches to (E)SEE discussed above and, in particular, is inconsistent with deflationary linguistic accounts, which treat (E)SEE as a linguistic default that should not be moderated by the type of entity involved in the scenario. As we will argue, taken together, these results indicate that the (Epistemic) Side-Effect Effect is not reducible to any single cognitive or linguistic mechanism but is instead a multifactorial phenomenon. The persistence of (E)SEE for unaware agents undermines EA and AA models, while the inverse (E)SEE observed in non-human cases contradicts LA explanations. These findings call for a reassessment of current theoretical models and suggest that future research should focus on exploring boundary cases to understand better the mechanisms underlying SEE and ESEE.