DNP 805 Topic5 Discussion Question Two
DNP 805 Topic5 Discussion Question Two
Using the clinical question you identified from above, determine the individual components to that question and pinpoint the location in the hypothetical database where the information you require will be extracted.
Hypothetical queries are queries embedding hypotheses about the database. The embedded hypothesis in a hypothetical query indicates, so to say, a state of the database intended for the rest of the query. Thus the answer to a hypothetical query h > q, with a hypothesis h, is in principle the result of evaluating q against the database revised with h. In case h is inconsistent with the database, query evaluation becomes a special case of counterfactual reasoning. However, the possible worlds semantics usually applied for this notion is not relevant for database applications due to reasons of inefficiency.
In this paper we discuss and compare different approaches to hypothetical queries, paying special attention to potentials for efficient evaluation. As a central part of the paper we present and discuss our own approach “counterfactual exceptions”, which have the important property of, as opposed to the other approaches discussed, requiring only minor overhead in query evaluation. This approach is thus realistic for practical implementation and use in environments supporting large databases. The “price” for efficient evaluation is an altered semantics, as compared to the other approaches. However, it can be argued that this semantics is at least as appropriate for database applications as that of the other approaches mentioned.
For knowledge base applications, assuming a small size and complex knowledge base, it may be reasonable to adapt a model-checking approach to counterfactual reasoning as that of Grahne and Mendelzon in [14]. For database
applications with a large size base and with efficient query-answering as an ultimate feature, it is unrealistic even in optimized implementations to consider all
close possible worlds, cf. the complexity results of [3, 4].
In order to seriously include hypothetical statements in a database query
language, we find it essential that the total amount of time spent on the evaluation of a hypothetical query is comparable to that of a conventional query.
This
makes it irrelevant to suggest a semantics which fully complies with a possible
DNP 805 Topic5 Discussion Question Two
worlds semantics.
In this paper, we compare different approaches to counterfactual reasoning, including our own proposal for a query evaluation mechanism intended for
databases [1]. We consider here only a single new “world” which adapts the hypothesis and we do not need to change the database at all, instead we modify
the query evaluation procedure so it behaves as if the relevant changes had been
made.
One important issue in our discussion concerns to what extent the result of
a hypothetical query should reflect the actual structure of the database. The
structure of the database is the means by which the knowledge in the database
is represented in facts, rules and integrity constraints, thus the question is to
what extend a semantics should go beyond a purely model-theoretical interpretation of the database.
We argue that hypothetical querying and counterfactual
reasoning in general, in many cases should exploit more than just the modeltheoretical consequences of the database. The actual database structure, especially the database rules representing the “why’s”, are also of concern, when assimilating hypothesis. To distinguish on this matter we introduce a logical property called structure independence, indicating that the structure is ignored. Our
mechanism takes into account more database structure than other approaches
to counterfactual reasoning, in a way which seems intuitively acceptable for
database applications.
We consider databases consisting of positive rules and facts together with integrity constraints that limit the possible updates. We define a deductive database
as a finite set of clauses of the form.