Aurélien Allard - Theory building and replicability: on the value of basic facts without theoretical foundations - Perspectives on Scientific Error 2024
For slides, see osf.io/ayfek/
The last 10 years have seen increased attention paid to the reproducibility of scientific experiments. This focus on reproducibility has met some pushback. One particularly interesting backlash from a philosophical point of view has concerned a group of psychologists and cognitive scientists who have promoted the superiority of theoretical concerns over pure replicability concerns (Buzbas & Devezer, 2023; Devezer et al., 2019, 2021; Feest, 2023; Flis, 2022; Haig, 2022; van Rooij & Baggio, 2020). According to these theory proponents, the focus on replicability is misguided, and risks to backfire if it is not supplemented or replaced with increased attention towards theory building.
Theory-reformers have put forward two main arguments in favor of focusing on theory-building, rather than on promoting reproducibility. The first argument is based on the premise that the value of replicability is dependent on theoretical sophistication. According to this idea, replicability without theory is of little value (Buzbas & Devezer, 2023; Devezer et al., 2021; Muthukrishna & Henrich, 2019). One major reason behind the no-value-without-theory argument relies on the idea that it is hard or even impossible to interpret experimental results without the underpinning of a proper theory, since theories are necessary to identify experimental effects. I call this idea the identification argument.
The no-value-without-theory argument is reinforced by the theory-as-means-towards-replicability argument (Muthukrishna & Henrich, 2019). According to this second argument, focusing on replicability here and now would be a bad idea even if replicability was our sole aim. In this framework, improved theory can be seen as both the ultimate goal of science, and as a means towards replicability. This second argument takes place within a global opportunity cost argument: there are more urgent issues to address than direct replicability issues, and scientists should prioritize other areas than purely methodological concerns.
This talk includes both a positive and negative contribution. On the negative side, I show the limits of the arguments promoted by theory reformers. On the positive side, I provide a general framework to understand the link between theory-building and the promotion of reproducibility.
I begin by examining the normative value of both building theory and establishing a-theoretical facts. Due to virtues such as simplicity, applicability, and breadth, I defend the major importance of theory building in science. However, this general high value of theories does not preclude the importance of establishing basic facts, especially if these facts contradict common beliefs among the scientific community.
Second, I examine the identification argument, and find it lacking. While some underlying assumptions are indeed necessary for inference, psychologists seem to be able to identify factors that are useful to identify experiments and effects.
Third, I examine the opportunity cost argument, and argue that, contrary to the assumptions of theory-reformers, opportunity costs generally favor establishing reproducibility practices as a means of promoting theory-building. While theory building can in some contexts contribute to reproducibility, this influence is not strong enough that it should preclude efforts at improving reproducibility on its own. Overall, improving reproducibility and improving theory should be seen as mutually reinforcing.
Негізгі бет Ғылым және технология Aurélien Allard - Theory building and replicability - PoSE 2024
Пікірлер