Argumentation and Information Situated Conviction Correction: On the Two-Sided Nature of Epistemic Change.


160 views
Uploaded on:
Description
Argumentation and Information Situated Conviction Amendment: On the Two-Sided Nature of Epistemic Change Fabio Paglieri CMNA IV College of Siena, Italy August 23-24, 2004 Cristiano Castelfranchi ISTC-CNR Roma, Italy Valencia, Spain Presentation: Conviction Correction and Argumentation
Transcripts
Slide 1

Argumentation and Data-Oriented Belief Revision: On the Two-Sided Nature of Epistemic Change Fabio Paglieri CMNA IV University of Siena, Italy August 23-24, 2004 Cristiano Castelfranchi ISTC-CNR Roma, Italy Valencia, Spain

Slide 2

Introduction: Belief Revision and Argumentation 1 st hidden case: Belief modification ( BR ) and argumentation procedures ( Arg ) are best comprehended/ought to be considered inside of the same calculated structure . Conviction correction = the route in which a specialists alters its own opinion , i.e. its own convictions. Argumentation = the route in which a specialists changes other agents’ mind , by impacting their convictions through correspondence. Two sides ( intellectual and social ) of the same epistemic coin. Related works : conviction change and correspondence (Galliers 1992) , conviction modification and defeasible thinking (Pollock, Gillies 2000; Falappa, Kern-Isberner, Simari 2002) . In CMNA IV : conviction systems for demonstrating argumentation (Carofiglio) , the part of convictions, objectives, and convictions over objectives in argumentation (Amgoud, Prade) . 2 nd fundamental case: keeping in mind the end goal to catch and model Arg , we require BR formalisms with a legitimate level of auxiliary intricacy . Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 3

Introducing Data-Oriented Belief Revision (DBR) Our model depends on a central qualification in the middle of information and convictions . Information will be data accessible to the specialists (i.e. accumulated and put away in his psyche), without and before any dedication to their unwavering quality. Convictions are those information that the specialists acknowledges as solid bases for activity, choice and particular thinking errands , e.g. deduction, expectation and clarification. In this system, information are chosen (i.e. either acknowledged or rejected as convictions) on the ground of their educational properties . Hence, changes after some time in the determination\'s results procedure focus conviction update: as it were, BR is a rising impact of information control . We call this procedure Data-situated Belief Revision ( D B R ) . Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 4

Introducing Data-Oriented Belief Revision (DBR) DBR depends on a theoretical model of epistemic handling significantly more mind boggling than the first AGM plan (Paglieri 2004) : Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 5

Focusing Belief Selection Introducing Data-Oriented Belief Revision (DBR) Informational properties of information (Castelfranchi 1996; Paglieri 2004) are: Relevance : a down to business\' measure utility of the datum, i.e. number and estimations of the (sought after) objectives for which the datum is required/helpful ; Credibility : a number\'s measure and estimations of all supporting information , diverged from all clashing information , down to outside and inner sources; Importance : a measure of the epistemic availability of the datum, i.e. number and estimations of the information that the specialists will be compelled to reexamine , if he reevaluate that solitary one; Likeability : a measure of the motivational request of the datum, i.e. number and estimations of the (sought after) objectives that are straightforwardly satisfied by that datum . These are social properties: in this manner, in DBR information are composed in systems , called Data Structures . Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 6

Tolerant full realist Prudent liberal realist Wishful deduction specialists Condition c f c f c f/(1 – l f ) Threshold k 0.5 0.6 0.8 Function F c f (c f + i f + l f )/3 c f ï‚\' (i f + l f ) Introducing Data-Oriented Belief Revision (DBR) Belief choice : dynamic information (i.e. information hopeful as convictions) are either acknowledged or rejected as convictions on the premise of their believability and/or significance and/or affability , depending by the determination parameters of that specific operators (i.e. the instructive properties that he considers most urgent in evaluating the unwavering quality of a given datum). Conviction determination in DBR is performed by a scientific framework shaped by a condition C , a limit k and a capacity F . C and k together figure out if the datum is acknowledged or not as conviction, while F relegates an estimation of quality to the relating conviction (if any). Given a datum f with validity c f , significance i f and affability l f , let B speak to the agent’s conviction set and B s f the conviction f with quality s . At that point the general type of conviction choice is: If C(c f , i f , l f ) ≤ k then B s f  B If C(c f , i f , l f ) > k then B s f  B with s f = F(c f , i f , l f ) Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 7

Support : f bolsters y (in images: f  y ) iff c y  c f , the validity y of is specifically relative to the believability of f .   Contrast : f contrasts y (in images: f  y ) iff c y  1/c f , the believability y of is on the other hand corresponding to the validity of f .   Union : f and y are united (in images: f & y ) iff c f and c y , mutually (yet not independently) focus the believability of another datum.   Introducing Data-Oriented Belief Revision (DBR) Data structures comprise in information (hubs) connected together by trademark relations (join). We characterize three fundamental relations: Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 8

S a R an a S g d g b R g S b R b Introducing Data-Oriented Belief Revision (DBR) Example : Rhett was mindful that his cherished Scarlett should take the US637 flight from Atlanta to New Orleans today, to visit her elderly wet medical caretaker Mamy. Viewing the news, Rhett is educated that there has been an appalling accident amid the arrival of that plane and all travelers passed on. Unglued, Rhett calls Mamy, who instructs him to take a few to get back some composure of himself and quit gabbing, subsequent to Scarlett arrived protected and sound at her home two hours prior. Information: a = Scarlett was on the flight US637 today b = all travelers of today flight US637 passed on in a crash g = Scarlett is at this moment at Mamy’s home, safe and sound d = Scarlett is dead RELATIONS: {( a & b )  d , g  d } Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 9

Modeling Argumentation in DBR Representation results in Data Structures: argumentation through credibility ; Toulmin’s model of contention; defeasible thinking . Expressivity results in DBR: [ join ] inconsistency administration; neighborhood versus worldwide argumentation methodologies. Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 10

a b f g d an e b f g d e Modeling Argumentation in DBR An essential component in argumentation is believability , i.e. how much the arguer\'s case fit in with the prior convictions of the gathering of people . In DBR, believability construct contentions work with respect to the datum\'s significance that the arguer needs to shield. Significance of the datum can be controlled utilizing two distinctive subjective procedures : Self-clear datum : the new datum is exhibited as taking after from what the group of onlookers definitely knew – the datum has not yet been construed, but rather it may have been, and the gathering of people is prone to comment: «Sure! Obviously! Obviously!⻠and so forth. Logical datum : the new datum is exhibited as supporting and disclosing information officially accessible to the group of onlookers – since such clarification was missing in this way, it produces responses like: «Now I see! That’s why! I knew it!⻠and so on. Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 11

D = information C = claim W = warrant Q = qualifier R = reply B = sponsorship Modeling Argumentation in DBR Toulmin’s model can be effectively spoken to in DBR as an impossible to miss Data Structure , as takes after: Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 12

C Direct defeaters D W Premise defeaters Undercutting defeaters Modeling Argumentation in DBR In DBR is it conceivable to speak to three unique sorts of defeaters : direct (refuting) defeaters = information differentiating the C hub; premise defeaters = information differentiating the D hub; undermining defeaters = information differentiating the W hub (i.e. counters). Case : John is blameless of the homicide of his wife ( claim ) in light of the fact that he cherished her much ( information ) and more often than not ( qualifier ) individuals don\'t kill the ones they adore ( warrant ), since homicide infers despise towards the casualty ( backing ). Direct defeater : «John had been seen shooting his wifeâ». Premise defeater : «John had a mystery affaire with another womanâ». Undermining defeater : «Jealousy can make you execute the ones you cherish mostâ». Paglieri, Castelfranchi: Argumentation and Data-Oriented Belief Revision CMNA IV, August 23-24, 2004, Valencia, ES

Slide 13

Modeling Argumentation in DBR While AGM ways to deal with BR bar inconsistencies on a basic level from the agent’s conviction set, argumentation ended up being an effective device for taking care of disagreements (e.g. Amgoud, Cayrol 2002; De Rosis et al. 2000; Pollock, Gillies 2000) . In DBR we recognize three sorts of commonly clashing data: information contrast : this is not in the slightest degree hazardous, yet rather an ad

Recommended
View more...