Coherence and Reliability: Studies in Bayesian Epistemology
(2011) Abstract
 In this thesis the connection between coherence and reliability is investigated. The question may be phrased as follows: does the fact that a set of testimonies is coherent imply that the witnesses who have reported these testimonies are reliable? The same question may also be expressed in terms of beliefs: does the fact that a set of beliefs is coherent imply that the beliefs were reliably produced? Traditionally, coherence theorists have thought that coherence is connected to truth, but in this dissertation the thesis is that it is rather connected to reliability.
The investigation proceeds within a probabilistic framework, the socalled witness scenario, where a number of partially reliable witnesses give independent... (More)  In this thesis the connection between coherence and reliability is investigated. The question may be phrased as follows: does the fact that a set of testimonies is coherent imply that the witnesses who have reported these testimonies are reliable? The same question may also be expressed in terms of beliefs: does the fact that a set of beliefs is coherent imply that the beliefs were reliably produced? Traditionally, coherence theorists have thought that coherence is connected to truth, but in this dissertation the thesis is that it is rather connected to reliability.
The investigation proceeds within a probabilistic framework, the socalled witness scenario, where a number of partially reliable witnesses give independent reports. Hence it is primarily coherent sets of testimonies that are discussed, but coherence of beliefs is taken to function similarly, and thus the results acquired using the witness scenario are, it is argued, relevant for the traditional discussion which mostly concerns coherence of beliefs. Using this scenario a notion of reliabilityconduciveness, similar to the muchdiscussed notion of truthconducivenes, is formally defined. This definition says, roughly, that a measure of coherence C is reliabilityconducive if and only if the more coherent a set of testimonies is, as measured by C, the higher is the probability that a witness who has given one of these testimonies is reliable. In Papers I and II it is tested whether a number of coherence measures proposed in the literature are reliabilityconducive in various salient scenarios. It is shown that the only coherence measure that is reliabilityconducive in all of those scenarios is the socalled Shogenji measure. This observation is then used to argue that the Shogenji measure is a fruitful explication of the notion of coherence. However, in Paper III it is shown that in the most general scenario neither the Shogenji measure nor any other measure of coherence is reliabilityconducive.
In Paper IV, the use of coherence reasoning in court is investigated. It is shown that we often use the notion of coherence in judicial reasoning, for example to find out whether certain witnesses are reliable or not. Using the results from Papers I and II, as well as additional formal results that strengthen the connection between (Shogenji) coherence and reliability, it is argued that this line of reasoning is valid. Also, it is argued in the introductory chapter that these findings show that coherence is indeed closely connected to reliability, even though it is not generally reliabilityconducive, and that this connection is much closer than that between coherence and truth.
In Paper V, coherent sets of socalled higher order testimonies (and higher order beliefs) are examined, where a higher order testimony is defined as a testimony concerning some other witness’s reliability (higher order beliefs are defined analogously). Such sets have hitherto not been examined within formal coherence theory. It is shown that coherence must be defined in a nonstandard way to account for all our intuitions regarding sets involving higher order testimonies. It is then argued that this shows that in order to correctly estimate the degree of coherence of a set of beliefs, the believer must have a firm grasp of her own reliability; a fact that in turn is taken to be problematic for the coherence theory of justification. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/2199486
 author
 Schubert, Stefan ^{LU}
 supervisor

 Erik J Olsson ^{LU}
 opponent

 professor Shogenji, Tomoji, Rhode Island College, Providence, USA
 organization
 publishing date
 2011
 type
 Thesis
 publication status
 published
 subject
 keywords
 coherence, Bayesian epistemology, reliability, higher order beliefs, reliabilityconduciveness, truthconduciveness, formal coherence theory, witness scenarios
 pages
 196 pages
 publisher
 Lund University (MediaTryck)
 defense location
 Sal 104, Kungshuset, Lundagård, Lund
 defense date
 20111126 10:15:00
 ISBN
 9789174731910
 language
 English
 LU publication?
 yes
 id
 4ac111d75a2d4d21b415c6c7351f2fcd (old id 2199486)
 date added to LUP
 20160404 10:27:35
 date last changed
 20181121 20:58:53
@phdthesis{4ac111d75a2d4d21b415c6c7351f2fcd, abstract = {In this thesis the connection between coherence and reliability is investigated. The question may be phrased as follows: does the fact that a set of testimonies is coherent imply that the witnesses who have reported these testimonies are reliable? The same question may also be expressed in terms of beliefs: does the fact that a set of beliefs is coherent imply that the beliefs were reliably produced? Traditionally, coherence theorists have thought that coherence is connected to truth, but in this dissertation the thesis is that it is rather connected to reliability.<br/><br> <br/><br> The investigation proceeds within a probabilistic framework, the socalled witness scenario, where a number of partially reliable witnesses give independent reports. Hence it is primarily coherent sets of testimonies that are discussed, but coherence of beliefs is taken to function similarly, and thus the results acquired using the witness scenario are, it is argued, relevant for the traditional discussion which mostly concerns coherence of beliefs. Using this scenario a notion of reliabilityconduciveness, similar to the muchdiscussed notion of truthconducivenes, is formally defined. This definition says, roughly, that a measure of coherence C is reliabilityconducive if and only if the more coherent a set of testimonies is, as measured by C, the higher is the probability that a witness who has given one of these testimonies is reliable. In Papers I and II it is tested whether a number of coherence measures proposed in the literature are reliabilityconducive in various salient scenarios. It is shown that the only coherence measure that is reliabilityconducive in all of those scenarios is the socalled Shogenji measure. This observation is then used to argue that the Shogenji measure is a fruitful explication of the notion of coherence. However, in Paper III it is shown that in the most general scenario neither the Shogenji measure nor any other measure of coherence is reliabilityconducive. <br/><br> <br/><br> In Paper IV, the use of coherence reasoning in court is investigated. It is shown that we often use the notion of coherence in judicial reasoning, for example to find out whether certain witnesses are reliable or not. Using the results from Papers I and II, as well as additional formal results that strengthen the connection between (Shogenji) coherence and reliability, it is argued that this line of reasoning is valid. Also, it is argued in the introductory chapter that these findings show that coherence is indeed closely connected to reliability, even though it is not generally reliabilityconducive, and that this connection is much closer than that between coherence and truth. <br/><br> <br/><br> In Paper V, coherent sets of socalled higher order testimonies (and higher order beliefs) are examined, where a higher order testimony is defined as a testimony concerning some other witness’s reliability (higher order beliefs are defined analogously). Such sets have hitherto not been examined within formal coherence theory. It is shown that coherence must be defined in a nonstandard way to account for all our intuitions regarding sets involving higher order testimonies. It is then argued that this shows that in order to correctly estimate the degree of coherence of a set of beliefs, the believer must have a firm grasp of her own reliability; a fact that in turn is taken to be problematic for the coherence theory of justification.}, author = {Schubert, Stefan}, isbn = {9789174731910}, language = {eng}, publisher = {Lund University (MediaTryck)}, school = {Lund University}, title = {Coherence and Reliability: Studies in Bayesian Epistemology}, year = {2011}, }