Cristian Larroulet Philippi

Measurement practices are central to most sciences. In the human sciences, however, it remains controversial whether the measurement of human attributes—depression, happiness, intelligence, etc.—has been successful. Are, say, widely used depression questionnaires valid measuring instruments? Can we trust self-reported happiness scales to deliver quantitative measurements as it is sometimes claimed? These and related questions are till today hotly disputed. There are two main frameworks under which human measurements are studied and criticized. One is the so-called construct validity framework. Here, criticisms to human measurements are typically of the form “this instrument is not valid: it does not actually measure the attribute we set out to measure”. The second framework is the standard typology of measurement scales (nominal, ordinal, interval, and ratio). Human measurements are commonly challenged for being merely ordinal—not quantitative: interval or ratio—despite the fact that many researchers use measurement results as if they were quantitative. The first part of the dissertation studies how adequate these frameworks are for evaluating measuring instruments and the inferences they afford. Regarding the concept of validity, it is commonly understood unconditionally, that is, without restricting validity judgments about measuring instruments to context-specific situations. Instruments are said to be (in)valid simpliciter. In Chapter 2, I argue against this conception of validity and in favor of a contextual one. Regarding the second main framework, the standard typology of scales is typically linked to a set of prescriptions regarding (un)justified measurement inferences. I call this classification-cum-prescriptions the “received view” on measurement scales. In chapters 3 and 4, I question the idea that the received view is an impeccable guide for clarifying the kinds of inferences we are licensed to make from measurement results in human science contexts. I motivate general doubts about the adequacy of the received view in Chapter 3, and I articulate these doubts in detail for the specific case of average group comparisons in Chapter 4. The upshot of this first part of the dissertation is a deeper awareness of the complexity surrounding which inferences can legitimately be made from measuring instruments. The first part of the dissertation is largely framed under the assumption that some human attributes are indeed quantitative, even if they are not currently being measured quantitatively. The second part addresses two important issues raised by that assumption. Chapter 5 tackles the so-called “quantity objection”: that human science attributes are themselves not quantitative, thus they cannot be quantitatively measured. This objection has been deployed to argue against optimistic positions regarding human quantification. I argue that the quantity objection is not successful in this sense—it begs the question to these optimistic human scientists that, just like their colleagues in the physical sciences have done, postulate theoretical quantitative attributes as working hypotheses. But what does treating human attributes quantitatively as working hypotheses amount to? And what are, or can be, “amounts” of depression, happiness, etc.? Chapter 6 argues that there is not one but various approaches for quantitatively conceiving of human attributes, each with its own success conditions. This chapter offers a conceptual framework for making progress on debates about controversial human measurements.

Wodak (2019) persuasively argues that we are not justified in believing that well-being measurements are linear. From this, he infers grave consequences for both political philosophy thought experiments and empirical psychological research. Here I argue that these consequences do not follow. Wodak’s challenges to the status of well-being measurements do not affect thought experiments, and well-being empirical researchers may be justified in making average comparisons even if their measurements are not linear.

There is a widely held view on measurement inferences, that goes back to Stevens’s ([1946]) theory of measurement scales and ‘permissible statistics’. This view defends the following prohibition: you should not make inferences from averages taken with ordinal scales (versus quantitative scales: interval or ratio). This prohibition is general—it applies to all ordinal scales—and it is sometimes endorsed without qualification. Adhering to it dramatically limits the research that the social and biomedical sciences can conduct. I provide a Bayesian analysis of this inferential problem, determining when measurements from ordinal scales can be used to confirm hypotheses about relative group averages. The prohibition, I conclude, cannot be upheld, even in a qualified sense. The beliefs needed to make average comparisons are less demanding than those appropriate for quantitative scales. I illustrate with the alleged paradigm ordinal scale, Mohs’ scale of mineral hardness, arguing that the literature has mischaracterized it.

While practitioners think highly of randomized studies, some philosophers argue that there is no epistemic reason to randomize. Here I show that their arguments do not entail their conclusion. Moreover, I provide novel reasons for randomizing in the context of interventional studies. The overall discussion provides a unified framework for assessing baseline balance, one that holds for interventional and observational studies alike. The upshot: practitioners’ strong preference for randomized studies can be defended in some cases, while still offering a nuanced approach to evidence-appraisal, one where not all non-randomized studies are treated equally.

There is a received view on measurement scales. It includes both a classification of scales and a set of prescriptions regarding measurement inferences. This article casts doubt on the adequacy of this received view. To do this, I propose an epistemic characterization of the ordinal/interval distinction, that is, one in terms of researchers’ beliefs. This novel characterization reveals the ordinal/interval distinction as too coarse grained and thus the received view as too restrictive of a framework for measurement research.

What is a valid measuring instrument? Recent philosophy has attended to logic of justification of measures, such as construct validation, but not to the question of what it means for an instrument to be a valid measure of a construct. A prominent approach grounds validity in the existence of a causal link between the attribute and its detectable manifestations. Some of its proponents claim that, therefore, validity does not depend on pragmatics and research context. In this paper, I cast doubt on the possibility of a context-independent causal account of validity. I assess several versions, arguing that all of them fail to judge the validity of measuring instruments correctly. Because different research purposes require different properties from measuring instruments, no account of validity succeeds without referring to the specific research purpose that creates the need for measurement in the first place.

Kitcher has proposed an ideal-theory account—well-ordered science (WOS)— of the collective good that science’s research agenda should promote. Against criticism regarding WOS’s action-guidance, Kitcher has advised critics not to confuse substantive ideals and the ways to arrive at them, and he has defended WOS as a necessary and useful ideal for science policy. I provide a distinction between two types of ideal-theories that helps clarifying WOS’s elusive nature. I use this distinction to argue that the action-guidance problem that WOS faces remains even under the aims/means distinction, because the WOS’s failure is more basic than critics have suggested.