Inconsistencies in Decision Modeling

NÃ¶hrer, Alexander

Inconsistencies in Decision Modeling

by Alexander NÃ¶hrer

Abstract:

Inconsistencies are a fact of life in software engineering. As a consequence, dealing and managing inconsistencies plays an important role in software engineering. However, while in some software engineering domains there exists support for living with inconsistencies, in the domain of decision-making they usually are disallowed â€“ either by preventing them or requiring them to be fixed right away. This stems mainly from the fact that common reasoning engines used in the decision-making domain do not handle inconsistencies well out of the box, in fact most of them fail to produce usable output after an inconsistency has been discovered. This thesis tackles the problem of managing and dealing with inconsistencies during decision-making. It presents two approaches, one that helps preventing inconsistencies by giving decision makers the maximum amount of freedom possible, while still providing meaningful guidance to reach oneâ€™s goals faster. The other approach presents different reasoning strategies for living with inconsistencies during decision-making, where one strategy in particular is very promising because it ensures that decision making automations remain working correctly even after an inconsistency is encountered (at the expense of completeness). While the focus of this thesis is primarily on the technologies â€œbehind the curtainsâ€�, it will also present one possibility of how to use and visualize these technologies in a prototype tool. Several evaluation scenarios will provide insight into the effectiveness of our approaches and their respective costs. It will show that our guidance calculation with respect to minimizing user input is nearly optimal and fast enough to be used interactively. It will also show that the cost of living with inconsistencies is acceptable, the level of incompleteness is reasonable and that living with inconsistencies can even have a positive impact on resolving inconsistencies. The key contributions of this thesis are: i) A unique perspective on the problems currently encountered during decision-making involving inconsistencies; ii) An approach that allows correct reasoning with SAT-Solvers in the presence of inconsistencies and automations to continue working without any adaptions; iii) An approach on how to determine the shortest path through a series of related decisions; iv) A basis for further research in general modeling scenarios; v) An extensive evaluation of the different aspects and techniques used in our approach.

Reference:

Alexander NÃ¶hrer, "Inconsistencies in Decision Modeling", PhD thesis, Johannes Kepler University (JKU), Linz, Austria, 2012.

Bibtex Entry:

@PhdThesis{Noehrer2012,
  Title                    = {Inconsistencies in Decision Modeling},
  Author                   = {Alexander NÃ¶hrer},
  School                   = {Johannes Kepler University (JKU), Linz, Austria},
  Year                     = {2012},

  Abstract                 = {Inconsistencies are a fact of life in software engineering. As a consequence, dealing
and managing inconsistencies plays an important role in software engineering. However,
while in some software engineering domains there exists support for living with
inconsistencies, in the domain of decision-making they usually are disallowed â€“ either
by preventing them or requiring them to be fixed right away. This stems mainly from
the fact that common reasoning engines used in the decision-making domain do not
handle inconsistencies well out of the box, in fact most of them fail to produce usable
output after an inconsistency has been discovered.
This thesis tackles the problem of managing and dealing with inconsistencies during
decision-making. It presents two approaches, one that helps preventing inconsistencies
by giving decision makers the maximum amount of freedom possible, while still providing
meaningful guidance to reach oneâ€™s goals faster. The other approach presents different
reasoning strategies for living with inconsistencies during decision-making, where
one strategy in particular is very promising because it ensures that decision making
automations remain working correctly even after an inconsistency is encountered (at
the expense of completeness).
While the focus of this thesis is primarily on the technologies â€œbehind the curtainsâ€�,
it will also present one possibility of how to use and visualize these technologies in a
prototype tool. Several evaluation scenarios will provide insight into the effectiveness
of our approaches and their respective costs. It will show that our guidance calculation
with respect to minimizing user input is nearly optimal and fast enough to be used
interactively. It will also show that the cost of living with inconsistencies is acceptable,
the level of incompleteness is reasonable and that living with inconsistencies can even
have a positive impact on resolving inconsistencies.
The key contributions of this thesis are: i) A unique perspective on the problems
currently encountered during decision-making involving inconsistencies; ii) An
approach that allows correct reasoning with SAT-Solvers in the presence of inconsistencies
and automations to continue working without any adaptions; iii) An approach
on how to determine the shortest path through a series of related decisions; iv) A basis
for further research in general modeling scenarios; v) An extensive evaluation of the
different aspects and techniques used in our approach.},
  File                     = {:PhD Theses\\2012 Alexander NÃ¶hrer\\NÃ¶hrer Alexander - Inconsistencies in Decision Modeling.pdf:PDF},
  Keywords                 = {FWF P21321-N15},
  Owner                    = {AK117794},
  Timestamp                = {2015.09.22}
}