1. Robots as a Context for Computer Science Education IPRE’s Approach The Institute for Personal Robots in Education Mark Guzdial, Georgia Institute of Technology Based on a talk by: Doug Blank, Bryn Mawr College with Tucker Balch, Deepak Kumar, Stewart Tansley, Jared Jackson, Natasha Eilbert, Keith O’Hara, Daniel Walker, Gaurav Gupta, Jay Summet, and Monica Sweat

2. IPRE Overview Research Project Mission: explore making CS education more fun and effective through the context of a personal robot Target: All levels, from middle school to graduate school Joint effort hosted at Georgia Tech with Bryn Mawr College (+ MSR)‏ 3 year seed funding provided by Microsoft Research (MSR)‏ Special ingredient and hypothesis: A personal robot for every student

3. IPRE: Lead Institutions Georgia Institute of Technology Tier 1 research university, founded in 1885 15,000 students 72% male All students required to take CS Bryn Mawr College Liberal arts college, founded in 1885 1,200 students Nearly 100% female Few students know they have CS

4. Hypotheses for the Challenges of CS Education Students don’t have a clear idea what CS is CS curricula have inherent and explicit biases that deter people from CS Current programming languages are overwhelming for beginners CS has been taught exactly backwards and upside down

5. IPRE Focus: A Personal Robot Every student gets their own robot Small enough to carry in backpack Cost about the price of a textbook Wireless, controlled from computer Interactive and easy to program Personalizable More than “just a robot”

6. Personal Robot turnLeft(.5)‏ speak(“Hello, Gamers!”)‏ playMusic(“madonna.wav”)‏ setFace(“smile”)‏ takePicture()‏ penDown(“red”)‏

7. IPRE’s Philosophy The Personal Robot provides the context The needs of the curriculum drive the design of the robot, software, and text The software should be easy to pick up, but scales with experience An accessible, engaging environment for new, diverse students Computer Science != programming Computing as a social activity Computing as a medium for creativity Focus on performances rather than competitions

8. IPRE Research Goals Hardware – robot Software – easy for programmers to write controllers Curricular Materials – focus on learning CS

9. Hardware: Ideal Personal Robot Medium for creativity Inexpensive Robust Fun

10. IPRE Pilot Hardware Kit Featuring Parallax’s Scribbler 6 Light sensors 7 IR sensors Stall sensor Speaker 5 LEDs 2 motors Bluetooth wireless Camera Gamepad

11. Scribbler with IPRE Fluke Now available at www.roboteducation.org

12. Serial Connection over Bluetooth USB Bluetooth Adapter Serial Bluetooth Adapter

13. Software: Goals Easy to learn, but doesn’t seem simplistic as the student grows in experience: “pedigogically scalable” Easy to use: no compile-download-run or other complications Instant gratification: interactive, dynamic Modern: be able to take advantage of existing and future robots Open source: available for study or change Cross-platform: core runs on Linux, Mac, and Windows

14. Python Looks like English Indentation matters Interactive Easy to learn, but powerful … IronPython gives access to Microsoft Robotics Studio … and Linux and Macintosh via Mono

15. Python Follow the Light def followLight(): setForwardness(0) left = 0 right = 2 while timeRemaining(30): if robot.getLight(left) < robot.getLight(right): turnLeft(1.0) if robot.getLight(right) < robot.getLight(left): turnRight(1.0) forward(1.0,0.1) followLight()

16. Myro: My Robot Library of functions Robot movements Sound and Music Communication Web interaction Vision and Image Processing Interfaces with Robotics Studio

17. Using Myro 1.Install software on any computer (Mac, Linux, or Windows) ‏ 2.Establish Bluetooth connection (via a Serial port) 3.Start Python 4.Load Myro 5.“Initialize” connection with robot 6.Have fun!

18. Start Python (IDLE) ‏

19. Load and Initialize connection >>> from myro import * >>> init(“com4”)‏ >>>

20. Testing Myro, without robot

21. Testing the Robot from myro import * init(“your COM port here”) forward(1,1) joyStick() beep(1,440)

22. Curriculum Goals Bring in examples from other related disciplines (e.g., biology, AI, humanities) ‏ Explicitly focus on robotics rather than programming constructs (e.g., chapter titles such as “Building Brains” rather than “Variables” or “Loops”) ‏ But, implicitly focus on Computing We believe this defines the notion of a context … and would work equally well with gameotics

23. Connections to Biology and Psychology

24. Programming as a social activity

25. Making stories with robots

26. “Civic Computing”

27. A CS1 Assignment: Exploring a Pyramid

28. 28 Robot Movies Cool project by Jay Summet: Creative, Collaborative – and Distributed/Parallel! Robots are characters. Multiple characters mean multiple students with multiple robots. One robot is camera How do you zoom? Aim and go forward! Challenges: How do you know when your actors are in their places? How do you “cue” the others? Post-processing media computation for eerie disappearing effects.

29. Example Movie

30. Games and Robots YouTube game videos available at cs.brynmawr.edu/games

31. Towards an Accessible, Engaging Environment for new, diverse students Competitions? Collaborations! Race? Orchestra! Compute factorial? Dance! Battlebots? A robot play! Draw a square? Abstract robot art!

32. Initial Assessment What was the most important or interesting thing that you learned in this course? "That computer science can be creative!"

33. Formative Interviews The robot did add a new dimension of excitement to the class. “It made it interesting to apply the computer programming to the robot – was not bland and gave it another dimension.” “Not many people can say 'yes I programmed a robot.' But now I can!” The robot was an additional complexity for the students. “Midway through we had tons of Bluetooth issues – I had to blindly write my code and then use someone else’s robot. Was unable to use mine for the last half of the semester and that was no fun.” “My robot died at that point but I would have done lots more than I was asked to do dancing, lights, music, etc.”

34. Formative Interviews It took effort to integrate the robot into the course. “[I] forgot [in lectures] that we were doing robots.” “We had one designated robotics TA for the whole class but he was only available to us twice a week. If homework is due and it's not time to talk to the TA, then we asked Monica and it was a lot for her. Sometimes the robotics TA didn’t know because it was new to him too.” “[It was] all robot in homework, but not in lecture.” Students were anxious about using the robot at first. “Thought it would be harder.” “[I was] scared of the robot.”

35. Assessment Results Three main trials so-far: Spring 2007: Attitudes robot (GT and Bryn Mawr) and non-robot (GT) ‏ Interviews to establish themes Surveys to test themes across whole class Fall 2007: More careful testing of learning, same groupings Spring 2008 vs. Spring 2009: Comparing similar cohorts, non-robots vs. robots

36. Attitudes in Spring 2007 All students enjoyed the robot, were comfortable with it, and found it easy to get working. Personalizing the robot improved the course, in students’ opinion. Reported that the class was about computer science Found homework challenging

37. Differences in Attitudes Spring 2007 BMC students did more on homework “because it was cool.” BMC students were undeclared majors. Reported being more excited about CS afterward. GT students were already declared majors. Less excited about robots overall, but more interested than BMC in more courses in computer science. Tended not to talk about the course to others.

38. Fall 2007: Final Exam Comparison at GT The final exam taken by all students had five shared questions. Shared questions did not require experience with the robot, but in some cases used “robotic” situations.

39. Example Question: Recursion

40. Results: Robot students did on average 10% better

41. Statistically Significant p <= 0.015 Ignore the Tracing Question

42. Confound: Differences in Class Demographics Due to the laptop requirement, advisors steered students who were declared as CS majors into the robots class, and other students into the non- robots class. 4% CS/Computation Majors in the Non-Robots class 81% CS/Computation majors in Instructor B's Robots class. (This is why you collect demographic information!) ‏

43. W’s vs. F’s, statistically significant

44. Using a Distributed Approach Several schools seeded with robots and funds, working with us on assessment. Shorter College Rowan University Georgia State University The University of Tennessee – Knoxville The University at Albany – SUNY Phillips Exeter Academy

45. Spring 2008 Non-Robots vs. Spring 2009 Robots Comparing robots vs. non-robots with demographics controlled: No difference in grade distribution No difference in pass/fail (WDF) rates Only 33% of enrolled students had prior knowledge that it was a robots class. Of those: 35% of students said that the robot was a positive influence on taking the course. 15% said it was a negative influence.

46. Robots vs. Non-Robots: Significant Attitude Differences

47. Differences in yes/no quests StatementSpring 2009 Robots Class Percentage Agreement Spring 2008 Non- Robots class My experiences in this class caused me to decide to take another computer science class. 27%32% I expect that I will have to write a program (in any language) after I finish this class. 41%39% During the class, I wrote a program that was not an assignment for this class. 36%27%

48. Status of Assessment Developing a standardized set of instruments Are not too hard or too easy Have little response to variations in instructor Are acceptable to a wide range of instructors Work with non-GaTech/BMC schools

49. Looking Forward Robotics Studio is too complex for our target audience, but we provide an easy path for students to take to explore advanced robotics through it Robotics Studio offers a Visual Programming Language that warrants exploration in the CS1 environment What about attraction and retention? Does the robot context help?

50. Bryn Mawr College Data for 12 years of CS2 CS2 Data Structures Enrollment

51. Looks Promising! CS2 Data Structures Enrollment

52. Future Work Implement our robot design Move from CPython to IronPython Pyjama CCLI Phase II Develop additional curricular resources

53. Conclusions Through the context of a Personal Robot we hope to make computing more personal, interesting, and effective Initial assessment shows that students learn about computing, find it challenging, yet enjoyable

54. Questions? For more information please visit cs.brynmawr.edu/games & www.roboteducation.org