Decide whether to allow uncertainty in user answers

During an interview, it is possible for attributes to all have a value of "uncertain". "Uncertain" is considered a valid attribute value, allowing the assessment to continue without requiring the user to answer every question.

An assessment can continue until the goal attribute has a value, which may itself be uncertain, but which may be able to return true or false based on other information provided by the user. If a definite answer to the investigated goal can be proved by some means, then it may be that the user does not have to obtain the missing information (eg where one option in a OR rule is uncertain, but another returns true).

In contrast, if the user could answer only 'yes' or 'no' or select a fixed value to every question, the assessment would stop until the user was able to answer the question.

Uncertainty and rulebase inheritance

In its most straightforward application, uncertainty reasoning operates to infer attributes to uncertain when the conditions that prove the attribute are uncertain. Consider the following hierarchy of rules:

attribute 1 is true if

attribute 2 is true

attribute 3 is true

attribute 4 is true

In an interview, if attribute 4 is set to uncertain, then that propagates through the rulebase in the following way:

attribute 3 is inferred to uncertain;

then attribute 2 is inferred to uncertain;

then attribute 1 is inferred to uncertain.

Higher-level attributes inherit uncertainty through inferencing. Unhandled, uncertainty can introduce unintended consequences in your applications. This idea applies equally to the 'unknown' operator.