overview

1. overview

2. switcharoo • a tool for rapid prototyping of physical interactive products • involves a collection of inexpensive, wireless physical input components that communicate with a Macromedia Director prototype on the screen • a model + switcheroo components + Director interface becomes a functioning physical prototype

3. Improving the Accuracy of Cell Phone Interruptions

4. The Effect of Contextual Information on the Behavior of Cell Phone Callers Daniel Avrahami, Darren Gergle, Scott Hudson, Sara Kiesler Carnegie Mellon University, HCII

5. The Goal • Daniel, please turn OFF your cell phone. • Thank you.

6. The Goal • The Problem: • The location-free nature of cell phones makes it hard for callers to predict receivers’ states, leading to inappropriate incoming calls.

7. The Goal • Possible Solution: • Providing callers with information about receivers’ context, to help them decide whether or not to call.

8. The Source of the Problem • How do callers decide when to call? • With Landline Phones they know: • The likely identity of the receiver • Their relationship with the receiver • The urgency of the call • Likely state of the receiver • Location of phone + Day + Time

9. The Source of the Problem (This isn’t good) ? ? ? • How do callers decide when to call? • With Landline Phones they know: • The likely identity of the receiver • Their relationship with the receiver • The urgency of the call • Likely state of the receiver • Location of phone + Day + Time Cell (This is good)

10. Possible Solutions • Single Rule • Manual Filtering • Multiple Rules (Profiles) • Automation • Caller-based • (Pedersen 2001; Schmidt, Takaluoma, Mäntyjärvi, 2000; Tang, Yankelovich, Begole, Van Kleek, Li and Bhalodia 2001).

11. The Goal • This Work: • Two experiments to asses the likely value of a caller-based solution. • We examined whether providing callers with information about receivers’ context improved the match between callers’ behavior and receivers’ desires.

12. Coming up… • Experiment 1 • Location and the presence of people • Experiment 2 (Suppa fast!) • Ringer status • Conclusions

13. Experiment 1 • 1 group played the role of Receivers • Receivers were asked to indicate whether they would prefer to receive a particular message in a set as • a phone call (immediate interruption) • by voice-mail (delayed interruption)

14. Experiment 1 • 4 groups played the role of Callers • Callers were asked to indicate whether they would prefer to deliver a particular message in the same set as • a phone call (immediate interruption) • by voice-mail (delayed interruption)

15. The Goal • Receivers: • You are in your office reading your email. A colleague is about to call you to tell you that John will be late for your meeting.

16. The Goal • Callers: • You are planning to contact a colleague on his cell phone to tell him that his favorite football team has lost. Your colleague is at home.

17. “Receiver” Condition

18. “Caller” Conditions

19. “Caller” Conditions • You are planning to contact a colleague on his cell phone to wish him a happy birthday.

20. “Caller” Conditions • You are planning to contact a colleague on his cell phone to tell him that his project got funding. Your colleague is at home.

21. “Caller” Conditions • You are planning to contact a colleague on her cell phone to tell her that she has a package in the mailroom. There are no people around your colleague.

22. “Caller” Conditions • You are planning to contact a colleague on his cell phone to tell him that his son is ill and is being sent home. Your colleague is at the office. There is at least one person around your colleague.

23. Messages • Your favorite football team has lost. • You have a movie overdue. • Your paper was accepted by the journal. • Your paper was rejected by the journal. • Happy Birthday! • Your project got funding. • You have a package in the mail room. • Your son is ill and is being sent home. • Your parents’ flight will be late. • Your next meeting is cancelled. • John will be late for your meeting. • We are at the coffee shop. You are welcome to join us. • Your car is ready for pickup. • Would you like us to buy you a ticket for the concert? • Your boss wants to talk to you. • Your assistant wants to schedule a meeting.

24. Urgency Ratings • All participants were asked to rate the urgency of each of the messages on a 5-point scale (1=low urgency, 5=high urgency)

25. Participants and Procedure • 78 Participants (26 Receivers and 52 Callers) • Average age 30.1 (SD=10.7, Min=18, Max=68) • Participants logged on to our website and were randomly assigned to condition. • Receivers chose for all the messages in all four conditions • Callers chose for each message once, each with randomly assigned level of contextual information • All participants assigned urgency ratings to all the messages

26. Experiment 1 – Results • Receivers’ behavior • Differences between Callers and Receivers • Effects of Contextual Information • Differences between conditions • Differences within conditions • Note: The data of 4 Callers were removed from the analyses as they chose to call 100% of the time.

27. Receivers’ behavior • Receivers’ Call Frequency (1=call, 0=voicemail) • Home (M=52.84%) vs. Office (M=44.55%) (p = .0001) • Alone (M=61.5%) vs. Not Alone (M=35.89%) (p < .0001) • Call Frequency by Location and People (p < .0001) • Call Frequency by Urgency (p < .0001)

28. Differences between Callers and Receivers • Receivers chose Call about 50% of the time, Callers chose Call about 60% (p = .002) • >> Without None condition (p = .003) • >> Call Frequency by Urgency for Callers (p < .0001) • >> Overall levels of Urgency did not vary across Role (M=3.14 vs. M=3.11, n.s.)

29. Effects of Contextual Information • Difference Score – The degree to which the Callers accurately anticipated Receivers’ preferences for a call (0=perfect, 1=perfectly wrong). • Calculated by subtracting from every Caller’s choice (1=call, 0=voicemail) the corresponding average Call frequency given by the Receivers for the same message in a matching situation (except for the None). EXAMPLE • The difference score for every contextual information condition was then computed as the absolute value of the average score by message.

30. Differences Between Conditions • Callers in the None condition were significantly worse(higher difference score) (p < .05) • No significant differences across the other contextual information conditions

31. Differences Within Conditions • Location Condition • Significantly more accurate when the receiver in the office than when at home (M=16.1% vs. M=20.2%, p < .0024) • Possible explanation: “office” helps to avoid interruptions

32. Differences Within Conditions • People Condition • Significantly more accurate when the receiver was alone than when others around (M=15.7% vs. M=23.7%, p < .0001) • Possible explanation: “alone” helps to interrupt appropriately

33. Differences Within Conditions • Combo Condition • Significantly more accurate when the receiver was alone (p < .0001) • Significantly more accurate when the receiver was at home (p < .0001) • Receiver at office and not alone, drastically inferior (28.9%) • Possible explanation: did not correct enough

34. The Goal • These findings suggest that providing different types of contextual information can result in different kinds of improvements in Callers’ accuracy.

35. Experiment 2 (do I have enough time?) • A new Caller condition – Cell Phone Ringer Status • Levels of the (Ringer) conditions were: • Loud • Silent (Flash and Vibrate) • “You are about to contact a colleague on her cell phone to tell her that you are at the coffee shop and she is welcome to join you. The ringer on your colleague’s cell phone is turned off (the phone will flash and vibrate).” • Everything else was the same…

36. Participants and Procedure • 12 Participants • Average age 27.7 (SD=6.08, Min=19, Max=43) • Callers chose for each message once, each with randomly assigned level of contextual information • All participants assigned urgency ratings to all the messages

37. Experiment 2 – Results • We used the full data from the Receiver condition from experiment 1. • Unlike Callers in experiment 1, role did not have a significant effect on the Call frequency of participants in the Ringer condition (M=50% vs. M=54%, n.s.)

38. Differences Between Conditions • We calculated difference scores similar to those calculated for Callers in the None condition • Callers in the Ringer condition were significantly better (lower difference scores) than Callers in the None condition (p < .0001)

39. Summary and Conclusions • This study tested the potential value of a Caller-based solution to the problem of inappropriate incoming cell-phone calls. • We believe that the relatively poor performance of Callers in the None condition compared to the other conditions confirms the existence of the problem.

40. Summary and Conclusions • Providing information improved Callers choices • Different information caused different kinds of improvements • Inherent differences between the roles of callers and receivers

41. Summary and Conclusions • Urgency plays a significant role for both Receivers and Callers and should be a significant factor when choosing between an automated solution and a caller-based solution) • Never ever use Comics Sans MS

42. Thank You