Please use this identifier to cite or link to this item: http://doi.org/10.25358/openscience-633
Full metadata record
DC FieldValueLanguage
dc.contributor.authorDolega, Krzysztof-
dc.date.accessioned2017-06-01T10:31:26Z-
dc.date.available2017-06-01T12:31:26Z-
dc.date.issued2017-
dc.identifier.urihttps://openscience.ub.uni-mainz.de/handle/20.500.12030/635-
dc.description.abstractIn this paper I discuss two leading positions regarding the representational commitments of the predictive processing approach to cognition. So called conservative predictive processing (cPP) (Clark2015a), claims that the explanatory power of the framework comes from postulating a rich nesting of genuine representational structures which come to serve as a model of the organism’s external and internal milieu (Hohwy2016; Gladziejewski2015a). Radical predictive processing (rPP), on the other hand, postulates that not all elements of the computational architecture should be interpreted as full-blown representations (Clark2015a). Instead, proponents of this approach argue that the relevant computational level descriptions (Marr1982) serve merely as abstract schemata aimed at capturing the dynamics of processes, which are embedded into neural structures by evolutionary selection (Orlandi2013; Orlandi2014; Orlandiforth) and do not depend on manipulating genuine representations (Clark2015b; DowneyThisCollection; BruinebergThisCollection). Surprisingly, both approaches are motivated by the adoption of William Ramsey's representational ‘job description challenge’ (Ramsey2007), according to which appeals to representational posits must be supported by a convincing demonstration of the relevant elements or structures playing a representational role within the functioning of a wider cognitive system. Gładziejewski Paweł Gładziejewski follows Ramsey in order to show that PP’s generative models perform a representational function by acting as detachable, information bearing structures which stand-in for the features of the environment, in order to enable capacities such as action-guidance and error detection. Clark Andy Clark and Orlandi Nico Orlandi, on the other hand, claim that only higher levels of the PP hierarchy exhibit such capacities, while lower (e.g. perceptual) levels should be construed as model-free structures governed by biases acquired through reinforcement learning or phylogenetic development. The main aim of this paper is to evaluate the competing approaches in relation to the underlying computational architecture and show that the disagreement between the two sides of the debate does not cut as deep as it has been presented in the literature. The main reason for this is that both sides agree on how to understand Ramsey’s representational challenge and the functionally distinct notions of representation it introduces. However, an investigation into the roles played by the structures posited by PP reveals that the job description challenge may be insufficient to differentiate the competing views. Although it does not support the radical claim that peripheral layers of PP systems consist in solely non-representational elements, it also poses a serious problem for the conservative side of the debate by inviting ambiguities with regard to the ascription of representational function to more nested structures. Proponents of cPP are tasked with distinguishing internal models targeting external, environmental features from structures functioning as meta-representations modelling the behavior of other parts of the system. I propose that the difference between the competing positions must come down to the question about the content of PP’s internal models. Supporters of rPP are faced with the task of introducing additional conditions which would help distinguish their position from the conservative one. Members of the cPP camp, on the other hand, can secure their interpretation by providing conditions for determining the contents of representational and meta-representational elements. Still, this is not an easy task due to the informational encapsulation of different layers of the system and the unclear conditions for identifying cases of misrepresentation. Because this paper is part of a larger collection beginning with a primer aimed at elucidating the main tenets of PP to an uninitiated audience (WieseMetzingerThisCollection), I will skip a typical introduction to the framework. Instead, I will start by articulating Ramsey's representational job description requirement for non-vacuous ascription of representational function (section 1 ), followed by a brief presentation of the two competing interpretations of the framework (section 2 ). Having delineated the available positions, I will move on to evaluate which of the elements posited by PP should be the target for the debate over the framework’s representational status (section 3.1 ). From there, I will argue that the representational job description challenge does not offer sufficient ground for distinguishing rPP from its conservative counterpart, by showing that it fails to secure a non-representational interpretation of the system’s peripheral layers (section 3.2 ). This, however, does not mean that cPP cannot be contested, as proponents of this reading must face the opposite problem of functional indeterminacy in models removed from the sensory periphery (sections 3.3 ). While, in principle, it is possible to resolve these issues by appealing to the contents of such posited representations, in practice this solution faces further difficulties, relating to a lack of clear conditions for content determination (section 3.4 ). I close the paper with a call for moderation in making claims regarding PP’s representational status, and point to two strategies for solving the problem at the heart of the debate (section 4 ).en_GB
dc.language.isoeng-
dc.rightsCC BY-NDde_DE
dc.rights.urihttps://creativecommons.org/licenses/by-nd/4.0/-
dc.subject.ddc100 Philosophiede_DE
dc.subject.ddc100 Philosophyen_GB
dc.titleModerate predictive processingen_GB
dc.typeBuchbeitragde_DE
dc.identifier.urnurn:nbn:de:hebis:77-publ-566525-
dc.identifier.doihttp://doi.org/10.25358/openscience-633-
jgu.type.dinitypebookPart-
jgu.type.versionPublished versionen_GB
jgu.type.resourceText-
jgu.organisation.departmentFB 05 Philosophie und Philologie-
jgu.organisation.number7920-
jgu.organisation.nameJohannes Gutenberg-Universität Mainz-
jgu.rights.accessrightsopenAccess-
jgu.book.titlePhilosophy and predictive processing-
jgu.book.editorMetzinger, Thomas-
jgu.pages.start161-
jgu.pages.end179-
jgu.publisher.year2017-
jgu.publisher.nameMIND Group-
jgu.publisher.placeFrankfurt am Main-
jgu.publisher.urihttp://dx.doi.org/10.15502/9783958573116-
jgu.organisation.placeMainz-
jgu.subject.ddccode100-
opus.date.accessioned2017-06-01T10:31:26Z-
opus.date.modified2017-06-02T07:53:51Z-
opus.date.available2017-06-01T12:31:26-
opus.subject.dfgcode00-000-
opus.organisation.stringFB 05: Philosophie und Philologie: Philosophisches Seminarde_DE
opus.identifier.opusid56652-
opus.relation.ispartofcollectionPhilosophy and predictive processingde_DE
opus.institute.number0508-
opus.metadataonlyfalse-
opus.type.contenttypeKeinede_DE
opus.type.contenttypeNoneen_GB
jgu.publisher.doi10.15502/9783958573116
jgu.organisation.rorhttps://ror.org/023b0x485
Appears in collections:JGU-Publikationen

Files in This Item:
  File Description SizeFormat
Thumbnail
56652.pdf400.89 kBAdobe PDFView/Open