Open courses are a sociocultural phenomenon. This phenomenon represents symptoms, responses, and failures facing Higher Education. In this talk, I examined open courses and MOOCs from a variety of angles and discussed the degree to which MOOCs have portrayed education as a product that can be packaged, automated, and delivered. Empirical research on the design and development of pedagogical and intelligent agents that may be used in MOOCs was also presented. More information here: http://www.veletsianos.com/2014/06/05/moocs-automation-artificial-intelligence-seminar/
MOOCs, Automation, Artificial Intelligence and Pedagogical Agents
1. MOOCs, Automation, Artificial Intelligence
and Pedagogical Agents
George Veletsianos, PhD
Canada Research Chair, Associate Professor
School of Education and Technology, Royal Roads University
June 18, 2014 :: University of Edinburgh
12. Today’s roadmap
1. The MOOC as a sociocultural phenomenon
2. Automation of teaching historically, and in the
context of MOOCs
3. Pedagogical agents as teaching automation
artifacts in online learning
13. - #change11 cMOOC
- MOOCs repurposed in my
courses (for student
analyses)
- Teaching an open course
in the Fall (Networked
Scholars - #scholar14)
- Research agenda
focuses on experiences
in emerging online
settings (e.g., open
courses, social media,
pedagogical agents)
14. - #change11 cMOOC
- MOOCs repurposed in my
courses (for student
analyses)
- Teaching an open course
in the Fall (Networked
Scholars - #scholar14)
- Research agenda
focuses on experiences
in emerging online
settings (e.g., open
courses, social media,
pedagogical agents)
15. The MOOC as a sociocultural
phenomenon
(with Rolin Moe)
16. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
17. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
• A symptom of the belief that education à workforce
training
18. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
• A symptom of the belief that education à workforce
training
• Representative of current political landscape
19. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
• A symptom of the belief that education à workforce
training
• Representative of the current political landscape
• Representative of the perspective that technology
provides solutions
20. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
• A symptom of the belief that education à workforce
training
• Representative of the current political landscape
• Representative of the perspective that technology
provides solutions
• Indicative of scholarly failures
21. If the MOOC is not just a learning model,
what is it?
• A response to increasing costs
• A symptom of the belief that education à workforce
training
• Representative of the current political landscape
• Representative of the perspective that technology
provides solutions
• Indicative of scholarly failures
• Representative of the belief that education can be
packaged and automated
23. An “industrial revolution” must occur in education, one
“in which educational science and the ingenuity of
educational technology combine to modernize the
grossly inefficient and clumsy procedures of
conventional education.”
Pressey (1933, pp. 582)
27. Basulto (2014) predicts
an “artificially intelligent machine” could teach massive
open online courses, “lecturing, grading and engaging
with students…Unlike humans, machines would be
willing to complete all the coursework and do all the
assignments…”
28. “With the menial job of checking and grading
assignments taken over by computers, we (human
teachers) will be left with the responsibility to intervene
and mentor our students” (Yair, 2014 in ACM Inroads)
The Automatic Teacher would free the teacher “from
mechanical tasks… so that she may be a real teacher,
not largely a clerical worker” (Pressey, 1927)
30. Margie’s future schoolroom in 2157 was
“right next to the bedroom, and the
mechanical teacher was on and waiting
for her… [all the] lessons were shown
and the questions were asked [on the
big screen].”
Asimov (1951)
32. Has the trend materialized?
“strong pressures to produce mediocre instructional
products based on templates and preexisting content”
(Wilson, Parrish, & Veletsianos, 2008, pp.42)
“Shovelware” = information masquerading as a course.
(Morrison & Anglin, 2006)
44. Tools à Functions
Email Scheduler à coordination
Study Group via OpenStudy à Pedagogical
Support
Interactive coding via Codeacademy à
Assessment
Ponti (in press)
45. These artifacts “remove the need for
exposure to teachers, by providing
participants with peer interactions and
automated coordination and testing”
(Ponti, in press)
46.
47. Automated courses: for learners that are
independent, self-organized, intrinsically
motivated and capable?
(Ponti in press; Tomkin & Charlevoix, 2014)
48. However, even though “MOOC teaching functions
are often disaggregated and delegated to
automated processes and community-based
social learning, the place and visibility of the
teacher remain of central importance.”
(Bayne & Ross, 2014)
49. Learners describe “a unique and powerful sphere
of intimacy that developed for them with their
xMOOC instructor, most especially in the context
of the pre-recorded instructional videos”
Adams, C., Yin, Y., Vargas Madriz, L. F., & Mullen, C. S. (in press). A
phenomenology of learning large: the tutorial sphere of xMOOC video lectures.
Distance Education.
50. The case of “Mary,” who wrote a short story
instead of an essay and shared the story on a
MOOC discussion board
51. “The professor was totally checked out, he never
visited the discussion board… and it was just
depressing and discouraging
And I thought his videos were not valuable at all so I
didn’t watch them. So, it was almost like that
course didn’t have an instructor …there was
someone who built the class and created the
reading but that was it.”
52. The artifacts in the Mechanical MOOC
reconfigured facilitation/instruction.
What other functions can automated artifacts
play?
62. Example #2: Agent appearance
Domagk (2010)
- Including an agent = no impact on learning (expected)
- Appealing agents promoted transfer
- Unappealing agents hindered learning
63. Agent-learner relationships &
agent-learner interactions
• A few studies in educational contexts - not
the majority
• These studies occur in open-ended
environments (not the norm in the field)
• Emerging evidence: enjoyment of social
chat, verbal abuse, fun with the system
• Computers As Social Actors (Media
equation), Uncanny valley
64. Research Questions
• What topics are discussed in agent-learner
conversations?
• What social practices emerge in agent-
learner conversations?
65. Series of studies
• Students have access to agents for weeks
at a time
• Naturalistic settings
• Variety of methods: Computer Mediated
Discourse Analysis, phenomenology, open
coding using standard interpretive lens,
quasi-experimental
66. Results #1: Small-talk
• Hey Mark, how are you today?
• Did you watch the [football] game last
night?
67. Results #2: Playfulness
• Did you watch the [football] game last
night?
• Do you have a girl/boyfriend?
68. Results #3: Abusive/aggressive
comments
• You stupid [expletive]!
• shut up. Don’t correct me.
• Agent: I can’t answer that.
User: WHY NOT!?
“The fact that he couldn't help me made
me really angry… I don’t remember what
the question was but [the agent] should
69. Conflicts
“I hated Joan or whatever the super-
agent lady was called. She asked me
at one point 'Are you testing me?' like
we were going to have some sort of a
confrontation or something. I've never
wanted to hurt a digital person
before!”
70. Results #4: Sharing personal
information
• I am worried about my exam score
• My girlfriend broke up with me
71. Results #5: Agent Role
• Agent as instructor/learning companion
(results from prior literature)
• Agent as mediator
– Can you tell professor X that she needs to
program you better?
• Agent as partner (sometimes you just want to
talk)
– This was an easy assignment, Mark.
72. • Social and psychological issues are as
significant as technology design issues.
• How would results differ:
– with different agents? (e.g., appearance)
– with agents of varied social intelligence?
– in MOOCs
– in studies of shorter/longer duration?