#4 Preferred Credulous reasoner fails

[mattsouth]

Via an email from Nikos:

"The preferred credulous reasoner using the enclosed informal argument source shows only A18 and A8 as green which I assume means that there exists a preferred extension that contains A18 and one that contains A8. But there are only two green arguments, so the possibilities is that there are two preferred extensions with one argument each, or one that contains them both. None however is possible because the resulting sets are clearly not admissible."

[mattsouth]

problematic graph - A8 and A18 cannot form an admissible set together or on their own

[mattsouth]

Following the presented example, I've managed to create a much smaller graph:
A4 A8
A8 A13 A14
A13 A4
A14 A18
A18 A13 A14
Which incorrectly suggests a single admissible set around A4.

This is a serious problem. The preferred credulous reasoner should not be used until it is fixed.

[mattsouth]

bumped up in priority by Xuetian Qi:

the PreferredCredulousReasoner's reasoning result is depended on the structure how to organize the attack relation ，for example ，i used the string "a c\n b a c\n c b\n d b c " to new a InformalArgumentSource,The unique preferred extension is {c},I changed the 135th line in the InformalArgumentSource.java(defeaters.put(argument, new HashSet<Argument<String>>())into (defeaters.put(argument, new TreeSet<Argument<String>>(new Comparator<Argument<String>>() {
public int compare(Argument<String> arg0, Argument<String> arg1) {
return arg1.getClaim().compareTo(arg0.getClaim());
}
}));),then the defeaters will be {a=[c], b=[c, a], c=[b], d=[c, b]},I wrote code to invoke the PreferredCredulousReasoner to query the c's acceptability ，then c is acceptable under preferred credulous semantics。
again ，if I changed the 135th line in the InformalArgumentSource.java into (defeaters.put(argument, new TreeSet<Argument<String>>(new Comparator<Argument<String>>() {
public int compare(Argument<String> arg0, Argument<String> arg1) {
return arg0.getClaim().compareTo(arg1.getClaim());
}
}));),then the defeaters will be {a=[c],b=[a, c], c=[b], d=[b, c]}, invoke the PreferredCredulousReasoner to query the c's acceptability ，then c is not acceptable under preferred credulous semantics。
so it shows that the argument's acceptability depends on the structure how to organize the attack relation，so your implementation of the PreferredCredulousReasoner maybe wrong ，because there're two different result for the same argument。