Ninja cursors use multiple synchronized cursors to reduce target acquisition time on large displays. An experiment with 4 cursor conditions (1, 2, 8, and 18 cursors) and 3 target densities found that 2 and 8 cursors performed best, reducing time by distributing targets across cursors. However, more cursors led to annoyance from frequent blocking and waiting. Advanced features like drag and drop and lasso selection were also explored.

1. Ninja Cursors Using Multiple Cursors to Assist Target Acquisition on Large Screens Masatomo Kobayashi (The University of Tokyo) Takeo Igarashi (The University of Tokyo)

2. Outline Background & Motivation Our Method Evaluation Discussion & Future Work Conclusion

3. Background Large display

4. Background Multi-display

5. Background (Virginia Tech) BlueSpace (IBM) Larger screens

6. Problem It is difficult to point to a distant object.

7. Introducing “ninja cursors” Demo

8. Basic idea of “ninja cursors”  The user can use the nearest cursor. Cover the screen with multiple, synchronously moving cursors.

9. Reducing the distance ( n : # of cursors) Average distance from the nearest cursor: n = 1 n = 4 D 

10. Studies on target pointing e.g., [Fitts 1954] Target Size Target Density e.g., [Guiard et al. 2004] + Cursor Size e.g., [Kabbash & Buxton 1995] + Cursor Density

11. Outline Background & Motivation Our Method Evaluation Discussion & Future Work Conclusion

12. Ambiguity problem What happens if multiple cursors point to multiple targets simultaneously?

13. Resolving ambiguity Only one cursor can point to a target; others are blocked and in the waiting queue. Queued Pointing Left Pointing

14. Resolving ambiguity Demo

15. Visual feedbacks Normal Pointing Blocked

16. Visual feedbacks Short waiting Long waiting Pointing

17. Outline Background & Motivation Our Method Evaluation Discussion & Future Work Conclusion

18. Goal Determine how the cursor number and the target density affect the performance.

19. Hypothesis # of Cursors Movement Time Effect of cursor blocking Effect of distance reduction

21. Setup

22. Cursor types 2 cursors 8 cursors 18 cursors 1 cursor (standard cursor)

23. Target numbers N = 1 N = 100 N = 400

24. Movement Time ( MT ) N = 2, 8 worked well. 0 0.5 1 1.5 2 2.5 (sec) 1 cursor 2 cursors 8 cursors 18 cursors N = 1 N = 100 N = 400

25. Error rate No significant trend. 0 1 2 3 4 5 6 7 N = 1 N = 100 N = 400 (%) 1 cursor 2 cursors 8 cursors 18 cursors

27. Outline Background & Motivation Our Method Evaluation Discussion & Future Work Conclusion

28. Advanced features Drag & drop Lasso tool

29. Drag & drop Drag Drop Drag with a cursor, drop with another cursor

30. Drag & drop Demo

31. Lasso tool Resolving ambiguity by implicit rules

32. Lasso tool 1. No lasso stroke ever intersects with targets. 2. Any lasso must contain at least one target. lasso not a lasso lasso not a lasso

33. Lasso tool Demo

34. Limitations Direct pointing devices cannot be used. Dense targets increase the MT too much.

35. Future work Combination with other techniques Measure the decision time

36. Decision time Cursors are visible even before each trial. Cursors are hidden until the start of each trial. vs. Compare 2 configuration:  Total Time = MT  Total Time = DT + MT

37. Decision time 0 0.2 0.4 0.6 0.8 1 1.2 1.4 MT DT+MT Total Time (s) 1 cursor 4 cursors

38. Regularly distributed targets

39. Regularly distributed targets

40. Regularity of cursors or targets ≈

41. Extra cursors or extra targets ≈

42. Bubbling ninja cursors

43. Bubbling ninja cursors 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Point Bubble Movement Time (s) 1 cursor 4 cursors

44. Post-selection method Use a post-selection menu instead of a waiting queue. + Does not increase the MT so much. + Does not modify the C-D gain.

45. Post-selection method Demo

46. Post-selection method Pie menu

47. Outline Background & Motivation Our Method Evaluation Discussion Conclusion

48. Related work Delphian Desktop [Asano et al. 2005]  Jump the cursor Bubble Cursor [Grossman & Balakrishnan 2005]  Change the cursor size Shadow Reaching [Shoemaker et al. 2007]  Use the shadow