1. Speaking while monitoring addressees for understanding Torsten Jachmann 16.12.2013 Herbert H. Clark and Meredyth A. Krych Seminar „Gaze as function of instructions - and vice versa“

2. Research Question Speaking and listening in dialog o Unilateral Speakers and listeners act autonomous No interaction o Bilateral Speakers and listeners monitor their respective partner Joint activity  What do speakers monitor?  How do they use that information?

3. Grounding Level 1 o Attend to vocalization Level 2 o Identify words, phrases and sentences Level 3 o Understand the meaning Level 4 o Consider answering

4. Grounding A: Where you there when they erected the new signs? B: Th… which new signs?(Level 3) A: Little notice boards, indicating where you had to go for everything B: No.  Bilateral account

5. Monitoring Voices o Attendance to partners utterances Faces o Gaze and facial expressions as indicator for understanding Workspaces o Region in front of the body o Manual gestures (but also games, etc.)

6. Monitoring Bodies o Head and torso movement as indicator Shared Scenes o Scenery beyond workspace Signals vs. Symptoms o Signals are constructed to get meaning across o Symptoms are not intentionally created

7. Least joint effort Opportunistic o Selection of the available methods that take the least effort to produce “Tailored” o Overhearers (not monitored by speaker) may misunderstand utterances

8. Method Pairs of directors and builders o 76 students (34 male / 42 female) Instructions to build 10 simple Lego Models 2 x 2 design (interactive) o 28 pairs Additional non-interactive condition o 10 pairs Video and audio analyses

9. Interactive Mixture model o Workspace (between subject) Visible Invisible o Faces (within subject) Visible Invisible No restrictions in time and talk

10. Non-interactive Only one condition Director records instructions o No time or talk constrains o Prototype can be examined as long as wanted before recording Builders listen to instructions o No constrains on actions Start, stop, rewind

11. Results Efficiency Turns Gestures and grounding o Deictic expressions o Gestures by addressees o Cross-timing of actions o Timing strategies o Visual monitoring

12. Efficiency Visibility of workspace improves efficiency

13. Efficiency Non-interactive Time needed to build much longer (245s “n-i” vs. 183s “i”) Strong drop in accuracy o Inadequate instructions

14. Turns Fewer SPOKEN turns of builder when workspace is visible

15. Deictic expressions Mainly unusable when workspace hidden o Joint attention needed o only referring to before mentioned situation

16. Gestures by addressees Mostly accompanied by deictic utterances (if any) Explicit verdict usually only on such utterances (otherwise continuing)

17. Cross-timing Gestural signals o Reflect understanding at that moment

18. Cross-timing Overlapping signals o Usually not in spoken dialog o Start with “sufficient information”

19. Cross-timing Projecting o Prediction of following actions/instructions

20. Cross-timing Initiation time o Waiting for partner to be able to attend the following utterance

21. Cross-timing Time uptake o Responses have to be timed exactly to the action and situation

22. Timing strategies Self-interruption o Dealing with evidence from the addressee o Usually not continued

23. Timing strategies Collaborative references o Deictic references rely on addressees actions

24. Visual monitoring Mainly used when director reaches a problem Eye gaze as support

25. Conclusion Grounding is fundamental Visible workspace enhances grounding speed In task-oriented dialogs faces are not important Compensation possible (only if any monitoring is available)

26. Conclusion Updating common ground Increments are determined jointly Much evidence for bilateral account o Addressees provide statement about current understanding o Speakers monitor to update and change utterances

27. Conclusion Opportunistic process o Offering options o Self-interruptions o Waiting o Instant revision Multi-modal process o Speech and gestures are combined if possible o Speech alone takes more time

28. Remarks Gaze only important for certain types of tasks Measurement of time maybe outdated (“old” study) No contradicting studies (To some extend commonsense)

29. Gaze and Turn-Taking Behavior in Casual Conversation Interactions Kristiina Jokinen, Hirohisa Furukawa, Masafumi Nishida and Seiichi Yamamoto

30. Differences Three-party dialogue No instructional task Stronger focus on eye gaze

31. Research Question How well can eye gaze help in predicting turn taking? What is the role of eye gaze when the speaker holds the turn? Is the role of eye gaze as important in three-party dialogs as in two-party dialogue?

32. Hypothesis In group discussions, eye gaze is important in turn to management (especially in turn holding cases) The speaker is more influential than the other partners in coordinating interactions (selects the next speaker)

33. Method Three-person conversational eye gaze corpus o Natural conversations o Balanced familiarity (50% familiar; 50% unfamiliar) o Balanced gender (male-only; female-only; mixed)

34. Method 28 conversations among Japanese students in their early 20’s with three participants each Each conversation about 10 minutes Eye gaze recorded for one participant

35. Method Eye tracker fixed on table to remain naturalness

36. Method

37. Used data Estimated at the last 300ms of an utterance if followed by a 500ms pause

38. Used data Dialog acts Speech features o Values of F0, etc. Eye gaze

39. Results

40. Conclusion Speaker signals whether he intends to give the turn or hold it by using eye gaze o fixating listener vs. focusing attention somewhere Eye gaze in multi-participant conversation as important as in two- participant conversations

41. Conclusion Eye gaze is used to select next speaker (seems to be correct) Maybe Japanese data interferes with value of speech data o Comparison Study? Listeners focus on speaker not vice versa

42. Remarks Vague information and data presentation o Although various data exists, interaction of factors is not presented o Some conclusions rely on the before mentioned point Setup only takes one participant in consideration Much of the data was unused o Lack in quality and way of creation

43. Remarks Study is based on data for another study o Setup is not optimal Realistic design o Yet, contains biasing flaws (situation of the participants, only one eye tracker)

44. Comparison Clark and Krych present interesting ideas but eye gaze is only rarely handled o How could this be altered? Jokinen et al. focus on eye gaze in a (more or less) natural situation but lack in scientific results and setup o What points and ideas of this setup could be beneficial?