<?xml version="1.0" standalone="yes"?>
<Paper uid="H86-1017">
  <Title>LIVING UP TO EXPECTATIONS: COMPUTING EXPERT RESPONSES'</Title>
  <Section position="5" start_page="185" end_page="185" type="concl">
    <SectionTitle>
5. Conclusion
</SectionTitle>
    <Paragraph position="0"> Since the behavior of expert systems will be interpreted in terms of the behavior users expect of cooperative human experts, we (as system designers} must understand such behavior patterns so as to implement them in our systems. If such systems are to be truly cooperative, it is not sufficient for them to be simply truthful. Additionally, they must be able to predict limited classes of false inferences that users might draw from dialogue with them and also to respond in a way to prevent those false inferences. The current enterprise is a small but non-trivial step in this direction. \[n addition to questions about achieving goals, we are investigating other cases where a cooperative expert should prevent false inferences by another agent, including preventing inappropriate default reasoning \[6, J~,VW84nonmon\].</Paragraph>
    <Paragraph position="1"> Future work should include s identification of additional eases where an expert must prevent false inferences by another agent, (r) formal statement of a general principle for constaining the search for possible false inferences, and s design of a natural language planning component to carry out the informing acts ~sumed in this paper.</Paragraph>
  </Section>
class="xml-element"></Paper>