Ponente: Philiphe van Basshuysen*
Fecha: 27 de abril, 12 hrs
Lugar: Sala de Investigadores Fernando Salmerón, Instituto de Investigaciones Filosóficas, UNAM.
Resumen: In a recent paper (2015), Dawid, Hartmann and Sprenger give a Bayesian argument in favour of the No Alternatives Argument: the conclusion from a lack of alternatives to a scientific theory to the truth of that theory. If valid, this would constitute a proof for the possibility of non-empirical theory confirmation. In this paper I argue that their "proof" begs the question.
I challenge the validity of the argument in two steps. I first give a principled line of reasoning concerning one critical assumption of their analysis, namely that the probability of a scientific hypothesis depends on the number of its alternative theories. I claim that arguing for this assumption implies either a regress to non-empirical theory confirmation, or a dubious application of the Principle of Indifference. In either case, the assumption stands on shaky grounds. In the second step, I give a counterexample to the assumption in order to show that it must indeed not be expected to hold.
I conclude with a more general claim. I claim that Bayesian analyses, and particularly the use of Bayesian nets, are “dangerous” in that they may entice one to be careless about implicit assumptions about probabilistic (in)dependencies, and how they support normative claims. I intend to broaden this investigation by more case studies and, finally, a general line of reasoning as to the boundaries of normative Bayesian analyses.
Lecturas previas:
Dawid, Hartmann y Sprenger
Basshuysen
Presentación.
*Debido a causas de fuerza mayor, la ponente original que presentaría en esta fecha (Ana Ponce) no podrá asistir al seminario y su charla quedará pendiente para un sesión posterior. Lamentamos los inconvenientes.
I challenge the validity of the argument in two steps. I first give a principled line of reasoning concerning one critical assumption of their analysis, namely that the probability of a scientific hypothesis depends on the number of its alternative theories. I claim that arguing for this assumption implies either a regress to non-empirical theory confirmation, or a dubious application of the Principle of Indifference. In either case, the assumption stands on shaky grounds. In the second step, I give a counterexample to the assumption in order to show that it must indeed not be expected to hold.
I conclude with a more general claim. I claim that Bayesian analyses, and particularly the use of Bayesian nets, are “dangerous” in that they may entice one to be careless about implicit assumptions about probabilistic (in)dependencies, and how they support normative claims. I intend to broaden this investigation by more case studies and, finally, a general line of reasoning as to the boundaries of normative Bayesian analyses.
Lecturas previas:
Dawid, Hartmann y Sprenger
Basshuysen
Presentación.
*Debido a causas de fuerza mayor, la ponente original que presentaría en esta fecha (Ana Ponce) no podrá asistir al seminario y su charla quedará pendiente para un sesión posterior. Lamentamos los inconvenientes.
No hay comentarios.:
Publicar un comentario