Michael Gelfond

opened his statement with reminding the purpose of a programming language, quoting C.A.R. Hoare:

``The main purpose of a programming language is to help the programmer in the practise of his art.''

and its influence, quoting E.W. Dijkstra:

``A ... programming language ... influences our thinking.''

He then proposed some principles for the design of declarative (programming) languages:

1. Take epistemology serious.

2. A new language feature should be accompanied by an in depth discussion of the methodology of its use.
3. Such a method can only be developed on the basis of a solid mathematical theory.

4. Use principles of the design of procedural languages.

As for item 1, Gelfond reminded that we represent knowledge about different items, objects and relations, e.g., hierarchy of classes, causal effects of actions etc. Under this perspective, the epistemological assumptions of ASP have to be critically reviewed. Knowledge about relations between objects is given by facts and rules, which together with the semantics assumption elicit further relations.

As for item 2, Gelfond thinks that the language methodology should provide a unique translation of the important natural language constructs into their formal counterparts. As an example, he discussed different translations of the statement ``A block can not occupy different locations.''

Regarding the questions of the moderator, on Q2 Gelfond said that KR has a main role, but that design of the language and implementation should be separated. On Q3, he said that the design of inference engines is important, and on Q7, that the next ASP language should provide

• universal quantifiers, of the form $a \leftarrow (\forall x p(x))q(x)$,
• efficient reasoning with numbers,
• reasonably efficient and general probabilistic reasoning, and
• learning.