Elamur Research⌕
☰
home / papers / 2006 / How to change a person's mind: Understanding the difference between the effects and consequences of speech acts

How to change a person's mind: Understanding the difference between the effects and consequences of speech acts

Paper · 2006 · Debora Field and Allan Ramsay

Summary

The paper tackles a core difficulty in planning communication. A speaker cannot directly change what a hearer believes. Whatever an utterance achieves depends on the hearer's own beliefs and inferences, so its outcome cannot be read off a fixed list of action effects. Field and Ramsay present a fully implemented planner, written in Prolog, that is at heart an epistemic theorem prover. It searches backwards from a goal through STRIPS-style operators and treats action effects as hypotheses inside proofs. This lets it reach goals that are entailed by an action's results rather than literally listed among them. Utterance meaning is planned with a single linguistic act and a set of Grice-derived rules about how hearers interpret maxim violations. Using the simple utterance type "There's a/an X!", the system models sarcasm, stating the obvious, bluffing and lying. It can also judge whether each would succeed or fail. The prover uses constructive logic, natural deduction and a deduction-based model of belief, which keeps reasoning about other minds tractable and closer to human inference.

Key points

  • The architecture interleaves a backward state-space search, using STRIPS operators, with a first-order theorem prover. Implications, disjunctions and negations can therefore be achieved as well as proved, and recursive rules are supported, such as above defined as the transitive closure of on.
  • A single linguistic act is used for all contexts. It is identified by surface form, with minimal preconditions and negligible built-in effects, which avoids the problems of the "speech acts with STRIPS" approach.
  • The sarcasm scenario is planned via a rule derived from Grice's first Quality maxim:
    • John says "There's an albatross!" about a pigeon.
    • Sally recognises an assertion they mutually believe is false.
    • She infers a flouted maxim, then humour, which leads to her being cheered up.
  • Deception is modelled from both sides:
    • Eight hearer violation (HV) maxims are implemented for detecting bluffs and lies, including a dodo example.
    • Speaker violation (SV) maxims cover four considerations: H's view of propositions, S's own view, S's perceived goodwill, and S's perceived reliability.
  • The prover is constructive and intuitionist and uses natural deduction. Belief follows a "deduction" model rather than possible worlds, so agents need not draw every logical consequence and need not have consistent beliefs.

Interesting ideas

  • I1. Instead of listing what a speech act causes, the system derives its consequences by proving what follows in the hearer's (modelled) belief state, so one utterance can yield different outcomes for different hearers.
  • I2. Sarcasm is detected by a chain: an assertion the hearer thinks both parties mutually believe is false signals a flouted Quality maxim, which then licenses a humorous reading; each step is an explicit rule.
  • I3. Nested belief is handled by switching into the believer's viewpoint and running ordinary first-order reasoning there, treating "H believes P" as "if I were H, I could prove P".
  • I4. Constructive implication rejects vacuous truths such as false-antecedent conditionals, so an implication holds only when a real derivation links the two sides.
  • I5. Backward search from goals is preferred because a speaker could say almost anything at any moment, which makes forward expansion hopeless.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): crate prag could keep speech-act frames minimal, deriving implicatures like irony or reassurance from rules over the speaker's and hearer's beliefs rather than hard-coding them per act type.
  • I2 (frames): an irony detector could check a statement against the world model's mutually known facts and emit a traceable "flouted Quality → non-literal intent" explanation instead of a classifier score.
  • I3 (world): the world of a text could store per-character belief sub-worlds and answer "what does the fox think?" by running the normal inference engine inside that sub-world.
  • I4 (answer): fable and test readers could require an actual proof path between premise and conclusion, giving explainable answers and refusing spurious ones.
  • I5 (back): generation from the meaning graph could start from the communicative goal and search backwards for content that entails it, rather than enumerating candidate sentences.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Speech act theory · 229 works

Cite

Cite as Debora Field and Allan Ramsay. 2006. How to change a person's mind: Understanding the difference between the effects and consequences of speech acts. Proceedings of the Fifth International Workshop on Inference in Computational Semantics (ICoS-5).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{field2006how,
  title = {{How to change a person's mind: Understanding the difference between the effects and consequences of speech acts}},
  author = {Debora Field and Allan Ramsay},
  booktitle = {Proceedings of the Fifth International Workshop on Inference in Computational Semantics (ICoS-5)},
  year = {2006},
  url = {https://aclanthology.org/W06-3903/}
}

References

29 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Allen, J. F. and C. R. Perrault, Analyzing intention in utterances (1980), AI 15: 143–78. find
  2. Appelt, D. E., Planning English referring expressions (1985), AI 26: 1–33. find
  3. Austin, J. L., How to do things with words (1962), Oxford: OUP, 2nd edition. find
  4. Blum, A. L. and M. L. Furst, Fast planning through planning graph analysis (1995), in Proc. 14th IJCAI, pp. 1636–1642. find
  5. Bonet, B. and H. Geffner, Heuristic Search Planner (2000), AI Magazine 21(2). find
  6. Bruce, B. C., Generation as a social action (1975), in B. L. Nash-Webber and R. C. Schank (eds), Theoretical issues in natural language processing, pp. 74–7. Cambridge, Massachusetts: ACL. find
  7. Bunt, H., Dialogue pragmatics and context specification (2000), find
  8. , pp. 81–150. [8] Bunt, H. and W. Black, (eds), Abduction, belief and context in dialogue: studies in computational pragmatics (2000), Philadelphia: John Benjamins. find
  9. Cohen, P. R. and H. J. Levesque, Rational interaction as the basis for communication (1990), find
  10. , pp. 221–55. [10] Cohen, P. R., J. Morgan and M. E. Pollack, (eds), Intentions in communication (1990), Cambridge, Massachusetts: MIT. find
  11. Cohen, P. R. and C. R. Perrault, Elements of a plan-based theory of speech acts (1979), Cognitive Science 3: 177–212. find
  12. Feigenbaum, E. A. and J. Feldman, Editors, Computers and thought (1995), Cambridge, Massachusetts: MIT Press. First published 1963 by McGraw-Hill. find
  13. Fikes, R. E. and N. J. Nilsson, STRIPS: A new approach to the application of theorem proving to problem solving (1971), AI 2: 189–208. find
  14. Green, C., Application of theorem proving to problem solving (1969), in Proc. 1st IJCAI, pp. 219–39. find
  15. Grice, H. P., Logic and conversation (1975), in P. Cole and J. Morgan, (eds), Syntax and semantics 3: Speech acts, pp. 41–58. New York: Academic Press. find
  16. Grosz, B. J. and C. L. Sidner, Plans for discourse (1990), [10], pp. 416–44. find
  17. Hintikka, J., Knowledge and belief: An introduction to the two notions (1962), New York: Cornell University Press. find
  18. Hoffmann, J. and B. Nebel, The FF planning system: Fast plan generation through heuristic search (2001), Journal of AI Research 14: 253–302. find
  19. Konolige, K., A deduction model of belief (1986), London: Pitman. find
  20. Kripke, S., Semantical considerations on modal logic (1963), in Acta Philosophica Fennica 16: 83–94. find
  21. Lewis, D., Scorekeeping in a language game (1979), J. Phil. Logic 8: 339–59. find
  22. McCarthy, J. and P. J. Hayes, Some philosophical problems from the standpoint of artificial intelligence (1969), Machine Intelligence 4: 463–502. find
  23. Newell, A., J. C. Shaw and H. A. Simon, Empirical explorations with the logic theory machine (1957), Proc. Western Joint Computer Conference, 15: 218–239. find
  24. Newell, A. and H. A. Simon, GPS, a program that simulates human thought (1963), [12], pp. 279–93. find
  25. Nguyen, X. and S. Kambhampati, Reviving partial order planning (2001), in Proc. IJCAI, pp. 459–66. find
  26. Pollack, M. E., Plans as complex mental attitudes (1990), [10], pp. 77–103. find
  27. Ramsay, A., Speech act theory and epistemic planning (2000), [8], pp. 293–310. find
  28. Searle, J. R., What is a speech act? (1965), in M. Black, (ed), Philosophy in America, pp. 221–39. Allen and Unwin. find
  29. Stalnaker, R., Pragmatics (1972), in D. Davidson and G. Harman, (eds), Semantics of natural language (Synthese Library, Vol. 40), pp. 380–97. Dordrecht, Holland: D. Reidel. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/field-2006-how-change-person-s-mind/