Borut Trpin

Explicating the concept of coherence and establishing a measure for assessing the coherence of an information set are two of the most important tasks of coherentist epistemology. To this end, several principles have been proposed to guide the specification of a measure of coherence. We depart from this prevailing path by challenging two well-established and prima facie plausible principles: Agreement and Dependence. Instead, we propose a new probabilistic measure of coherence that combines basic intuitions of both principles, but without strictly satisfying either of them. It is then shown that the new measure outperforms alternative measures in terms of its truth-tracking properties. We consider this feature to be central and argue that coherence matters because it is likely to be our best available guide to truth, at least when more direct evidence is unavailable.

Recent critiques have cast doubt on the viability of probabilistic measures of explanatory power. We respond by developing a coherence-based family of measures that sidesteps these challenges. Rather than assessing explanatory power solely by how well a hypothesis accounts for the evidence, these measures evaluate how well the entire explanatory package, including multiple explanantia, coheres with the explanandum compared to how well it coheres with its negation. We show that this approach accommodates key scientific judgments while preserving the virtues of a probabilistic framework.

Inference to the Best Explanation (IBE) is traditionally conceived of as a rule of inference, in which one infers to the hypothesis that provides the best explanation of one’s evidence. But what if some of that evidence is uncertain?How, if at all, can the traditional conception of IBE be extended to handle this common class of cases? This paper presents a new general model for investigating rules of inference from uncertain evidence, and then applies this approach to evaluate several different ways of extending IBE to handle uncertain evidence. The model employs computer simulations in which inference rules receive rewards (penalties) for making correct (incorrect) inferences in randomly generated scenarios in which the evidence is uncertain to various degrees. The model allows us to move beyond reliance on intuitive or historical examples in evaluating how well different inference rules handle cases in which the evidence is uncertain.

We investigate the epistemic role of coherence in scientific reasoning, focusing on its use as a heuristic for filtering evidence. Using a novel computational model based on Bayesian networks, we simulate agents who update their beliefs under varying levels of noise and bias. Some agents treat reductions in coherence as higher-order evidence and interpret such drops as signals that something has gone epistemically awry, even when the source of error is unclear. Our results show that this strategy can improve belief accuracy in noisy environments but tends to mislead when evidence is systematically biased. We explore the implications for the rationality of coherence-based reasoning in science.

What is the rational response to a scientific disagreement? Many epistemologists argue that disagreement with an epistemic peer should generally lead to conciliation by lowering confidence in the disputed belief or even suspending judgment altogether. Although this conciliatory approach is widely regarded as a norm of individual rationality, its value in the context of collective scientific inquiry is less clear. Some have even raised concerns that conciliating in scientific disagreements may slow progress or reduce the efficiency of inquiry. In this paper, we introduce a novel agent-based model that captures key aspects of scientific disagreement by incorporating both epistemic norms, which govern belief revision, and zetetic norms, which guide how scientists pursue inquiry. Our results indicate that the effects of conciliating in the face of disagreement---whether detrimental or beneficial---depend on the zetetic norms that scientists follow. When they focus on exploiting the hypothesis that they believe is most likely to succeed, remaining steadfast is more effective. However, with exploratory scientists, conciliation does not negatively affect group performance. These findings highlight the critical role of zetetic norms in determining the rational response to disagreement in scientific practice.

A conjunction of two hypotheses may provide a better explanation than either one of them individually, even if each already provides a good explanation on its own. An appropriate measure of explanatory power should reflect this, but none of the measures discussed in the literature do so because they only consider how much an explanatory hypothesis reduces our surprise at the evidence – which is problematic. This chapter introduces and defends a class of coherentist measures of explanatory power, and shows that one particular such measure that combines different intuitions underlying the epistemological notion of coherence overcomes the aforementioned problems of surprise reduction measures.

Alongside science and law, argumentation is also of central importance in everyday life. But what characterizes a good argument? This question has occupied philosophers and psychologists for centuries. The theory of Bayesian argumentation is particularly suitable for clarifying it, because it allows us to take into account in a natural way the role of uncertainty, which is central to much argumentation. Moreover, it offers the possibility of measuring the strength of an argument in probabilistic terms. One way to do this, implicit in much work, is to identify the strength of an argument with the degree to which the premises of the argument confirm the conclusion. We criticize this prima facie plausible proposal and suggest instead that the strength of an argument has something to do with how much the premises and the conclusion of the argument cohere with each other. This leads to a new probabilistic measure whose properties we examine in more detail.

Coherence considerations play an important role in science and in everyday reasoning. However, it is unclear what exactly is meant by coherence of information and why we prefer more coherent information over less coherent information. To answer these questions, we first explore how to explicate the dazzling notion of ``coherence'' and how to measure the coherence of an information set. To do so, we critique prima facie plausible proposals that incorporate normative principles such as ``Agreement'' or ``Dependence'' and then argue that the coherence of an information set is best understood as an indicator of the truth of the set under certain conditions. Using computer simulations, we then show that a new probabilistic measure of coherence that combines aspects of the two principles above, but without strictly satisfying either principle, performs particularly well in this regard.

Should a scientist rely on methodological triangulation? Heesen et al. (Synthese 196(8):3067–3081, 2019) recently provided a convincing affirmative answer. However, their approach requires belief gambles if the evidence is discordant. We instead propose epistemically modest triangulation (EMT), according to which one should withhold judgement in such cases. We show that for a scientist in a methodologically diffident situation the expected utility of EMT is greater than that of Heesen et al.’s (2019) triangulation or that of using a single method. We also show that EMT is more appropriate for increasing epistemic trust in science. In short: triangulate, but do not gamble with evidence.

Some of the information that we learn comes to us in a conditional form. This has proven to be a problem for philosophers, who try to explain how probabilistic beliefs change when one learns from conditional sentences. The problem is that a straight-forward solution is not possible: the partial belief in the antecedent and the partial belief in the consequent either increase, decrease, or remain the same. Two existing approaches to learning from indicative conditionals are considered: an explanatory one, and another that builds on relative information minimizing with regard to the causal structure. A novel method based on epistemic entrenchment is proposed to overcome the drawbacks of the competing approaches. The method solves all the standard examples and some other examples for which existing approaches have failed to provide adequate solutions.

Partial lying denotes the cases where we partially believe something to be false but nevertheless assert it with the intent to deceive the addressee. We investigate how the severity of partial lying may be determined and how partial lies can be classified. We also study how much epistemic damage an agent suffers depending on the level of trust that she invests in the liar and the severity of the lies she is told. Our analysis is based on the results from exploratory computer simulations of an arguably rational Bayesian agent who is trying to determine how biased a coin is while observing the coin tosses and listening to a liar’s misleading predictions about the outcomes. Our results provide an interesting testable hypothesis at the intersection of epistemology and ethics, namely that in the longer term partial lies lead to more epistemic damage than outright lies.

It has been argued that if the rigidity condition is satisfied, a rational agent operating with uncertain evidence should update her subjective probabilities by Jeffrey conditionalization or else a series of bets resulting in a sure loss could be made against her. We show, however, that even if the rigidity condition is satisfied, it is not always safe to update probability distributions by JC because there exist such sequences of non-misleading uncertain observations where it may be foreseen that an agent who updates her subjective probabilities by JC will end up nearly certain that a false hypothesis is true. We analyze the features of JC that lead to this problem, specify the conditions in which it arises and respond to potential objections.

This doctoral dissertation investigates what influence indicative conditionals have on belief updating and how learning from conditionals may be modelled in a probabilistic framework. Because the problem is related to the interpretation of conditionals, we first assess different semantics of indicative conditionals. We propose that conditionals should be taken as primary concepts. This allows us to defend a claim that learning a conditional is equivalent to learning that the relevant conditional probability is 1. This implies that learning a conditional can be modelled as learning a material conditional. We then turn to a few competing approaches to the problem of learning from a conditional -- an approach that models these situations with regards to changes in the explanatory status of the antecedent, as minimising probability distances between the prior and posterior probability distributions, and, particularly, an approach based on epistemic entrenchment. It is argued that although all approaches are important for their pioneering contributions, none of them ultimately provides a fully satisfactory solution. We then formulate a proposal according to which learning a conditional may be modelled as learning a material conditional in conjunction with other contextually inferred or observed information. We show that the view has an intuitive appeal on the grounds of pragmatics and empirical research in psychology of reasoning. We also show that the proposed approach successfully resolves all standard cases from the literature. The proposal is then extended to left-nested conditionals. To do so, we first inspect some preliminary empirical data about actual reasoning with nested conditionals and then show that the proposed approach also successfully models these cases. We then investigate how the approach may be extended to the so-called imperative conditionals. We introduce a tripartite typology of imperative conditionals and argue that although imperatives cannot be probabilistically modelled in any straight-forward way, at least some general insights of the proposed method help us understand the influence that the three different types of imperative conditionals have on listeners’ beliefs. We conclude with a summary and some open questions that should be addressed in future work (e.g. the influence of counterfactuals on belief updating).

Characters in Richard II utter a number of neglected, yet philosophically interesting imperative conditionals. Based on a close reading of these examples, I provide a tripartite typology of imperative conditionals. The type 1 constitute the class of standard imperative conditionals; the type 2 implicate that the antecedent is false; and the type 3 implicate that the command in the consequent is to be complied with. I show how the type 2 and type 3 conditionals can be identified, and explain when and why the speakers utter them.

It has recently been argued that a non-Bayesian probabilistic version of inference to the best explanation (IBE*) has a number of advantages over Bayesian conditionalization (Douven [2013]; Douven and Wenmackers [2017]). We investigate how IBE* could be generalized to uncertain evidential situations and formulate a novel updating rule IBE**. We then inspect how it performs in comparison to its Bayesian counterpart, Jeffrey conditionalization (JC), in a number of simulations where two agents, each updating by IBE** and JC, respectively, try to detect the bias of a coin while they are only partially certain what side the coin landed on. We show that IBE** more often prescribes high probability to the actual bias than JC. We also show that this happens considerably faster, that IBE** passes higher thresholds for high probability, and that it in general leads to more accurate probability distributions than JC.