Content uploaded by Orit Shaer
Author content
All content in this area was uploaded by Orit Shaer on Mar 31, 2020
Content may be subject to copyright.
Pre-Print for IEEE Computer Society
Teaching Pervasive Computing: A Report and a
Look Ahead From a Dagstuhl Seminar
Andrew L. Kun
University of New Hampshire
Anne Roudaut
University of Bristol
Audrey Girouard
Carleton University
Orit Shaer
Wellesley College
IN A RECENT Education & Training column, we addressed three central questions regarding
teaching pervasive computing: Why is it important to teach pervasive computing, What
should we teach students, and How should we do it?
1
Subsequently, we organized a
Dagstuhl Seminar entitled “Ubiquitous Computing Education: Why, What, and How” to
explore these questions in more detail.
2
The workshop gathered 26 faculty members and
one undergraduate student
3
to discuss the current state of ubiquitous computing
education and to provide ideas for how to improve on our current practices. In this column,
we discuss, and expand upon, the work of the seminar, including a detailed overview of the
challenges of teaching pervasive computing, proposing a curriculum for students with
different backgrounds, and exploring innovative active learning methods for pervasive
computing.
INTRODUCING DAGSTUHL:
A SEMINAR SERIES IN A GERMAN CASTLE
Before we report on the results of the seminar, let us first introduce Dagstuhl. Schloss
Dagstuhl—Leibniz Center for Informatics is the German organization that hosts the highly
prestigious Dagstuhl seminar series. Organizers propose seminars, and if the scientific
committee accepts the proposal, Dagstuhl takes on the logistics of the event and provides
financial support such that participants can attend at a low cost. The location is idyllic: it is a
castle (Figure 1) located in the south-west of Germany, where meeting rooms combine
modern amenities with old-world charm. The grounds offer many miles of scenic paths to
walk, jog, or bike, e.g., the ruins of a medieval fort are only a few minutes from the meeting
rooms. Participants share meals in the castle restaurant, and usually chat late into the night
in the pool room or the cellar. This place is made for engaging in deep discussions with
your colleagues.
Figure 1. Dagstuhl Castle is the location of tens of computer science seminars each year.
ON TEACHING PERVASIVE COMPUTING
We organized our workshop around the central themes we proposed in our recent
Education & Training column: Why is it important to teach pervasive computing, What
should constitute the curriculum, and How should we teach the curriculum? On day one,
following a round of introductions, we explored the question “why?” and we also discussed
the various challenges of teaching pervasive computing. Participants brainstormed ideas
around grand challenges. On day two, we explored the question of what we should teach
about pervasive computing. Participants organized small groups and worked to create a
curriculum for ubicomp education for students of various backgrounds (e.g., technical
versus humanities, and university versus industry), different degree levels (undergraduate,
graduate, or a short training program). We devoted days three and four to how we should
teach pervasive computing. Participants generated a list of their current active learning
2
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
methods or tools and exchanged them in a speed networking fashion with each other.
Participants also developed and experienced new active learning pedagogies on ubiquitous
computing topics. We also discussed pedagogies for academic ubiquitous computing
(ubicomp) programs. Finally, on day five, we wrapped up the seminar and formulated plans
for concrete collaborative actions.
Why?
Perhaps our most animated discussion was about the definition of pervasive (ubiquitous)
computing. What exactly is ubicomp? Clearly, ubicomp brings together multiple disciplines.
So, which disciplines, or subdisciplines, are at the core of any definition of the field? Is it
human– computer interaction? Networking? Embedded electronics? Our discussion made it
clear that there is no broad agreement on this topic, and there would be value in a detailed
description of the elements of this field.
While not everyone agreed on a single definition of ubicomp, we all shared the
Weiserinspired broad working definition that ubicomp is a field that explores connected
computing devices that are embedded in the fabric of our lives.
4
From this starting point,
we identified a number of issues that can help us determine why we should teach ubicomp.
One line of reasoning revolved around the idea that research and ubicomp are intimately
connected. This is because ubicomp remains a new and rapidly developing area. Thus,
teaching ubicomp can be a vehicle to instruct students in the general tools of creating new
knowledge: the tools of the scientific method. This could be especially valuable in academic
programs that historically focus on specific tools for generating new knowledge (such as
the tools of math, physics, and engineering).
Another interesting topic was the relationship between industry and academia. Where is
the balance between industry motivating the need for teaching ubicomp, and academia
providing possible paths forward in research? What are the roles of research publications
and products? Which is important, and when? How do the answers to all these questions
change over time? We did not reach conclusions, but our discussions clarified that the role
of industry in education and training is a hot topic that deserves a great deal of attention.
Another topic discussed in the seminar is the need for diversity of perspectives and
disciplinary backgrounds in the workforce. We agreed that one reason to teach ubicomp is
3
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
that our students need to become experts who are comfortable solving problems and
innovating in a multidisciplinary environment.
What?
So, what is “ubicomp education done right?” One aspect of this question is what content we
should teach. We first proposed that the topics to teach depend on the student audience as
follows.
1. Are we designing for students in a standalone course (e.g., a one-semester
undergraduate course for computer science students), or a program (e.g., a
graduate certificate program in ubicomp)?
2. What is the target education level? Are we designing for undergraduate students,
graduate students, employees of a tech company, or some other group?
3. What are the students’ backgrounds? Do they all have a technical background in a
single discipline (for example, electrical engineering)? Or do the students form a
multidisciplinary group, for example, with some having a technical background and
others with a social science background?
Next, we introduced a list of topics (grouped into themes) that would be of interest to
various students—either because these topics are already part of some courses related to
ubicomp, or because we feel they should be in the future. We then collaboratively created
a matrix of student characteristics versus topics.
2
This matrix can act as an inspiration for
those designing various teaching and training curricula for ubicomp. Here, we want to
point out two issues related to the matrix.
First, the list reflects those who created it. The majority of workshop participants had a
technical background and so most of the topics in the matrix are technical. The topics are
multidisciplinary, in the sense that they cover different technical topics from electrical
engineering, to computer engineering, to computer science, to mathematics. One
important nontechnical topic that is included in our matrix is ethics. We recognize that
ubicomp technology and products impact individuals, communities, and society at large,
and seek for our students to be able to carefully consider the intended and unintended
consequences of the products they envision and develop. But we expect that students will
4
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
benefit from curricula that go beyond a brief discussion of ethics, and include a variety of
both technical and nontechnical topics. For example, an exposure to topics in psychology,
sociology, and anthropology will benefit technically oriented students just as exposure to
coding and electronics prototyping will benefit those in the social sciences.
Second, it is instructive to see how topics of interest to ubicomp change over time. Our
matrix includes prototyping and fabrication techniques. In contrast to a decade ago, today
an instructor can include these topics in ubicomp training because the tools for prototyping
and fabrication are widely available at affordable prices. We expect that two technologies
that will soon be similarly widely available at reasonable prices are mobile eye tracking and
augmented reality. Today, devices in these two categories are available, but the cost is still
several thousand dollars per device, in contrast to tens of dollars for a prototyping board.
How?
Next, we focused on pedagogy—How should we teach students? Our approach was first to
list and discuss the issues that are directly or indirectly related to pedagogy, and second, to
generate new educational material that can complement existing educational resources.
One observation that emerged from our discussions is that instructors must make sure
that students can manage the workload of an interdisciplinary ubicomp course. Many
ubicomp courses present material related to a number of disciplines. Furthermore, many
of these courses have a project component, where students might need to be familiar with
topics from a variety of disciplines. Instructors in these courses could inadvertently
overwhelm students with the sheer number of topics that the course covers.
A related problem is that of deciding how much a student needs to learn about a topic.
When are we satisfied with students only achieving shallow understanding, and when do
we aim for them to have deeper insights? Furthermore, how does this vary from one
student to another in the same course? For example, if a ubicomp course project brings
together a psychologist and a computer scientist, how much do we want the psychologist
to learn about sensing, and how much do we want the computer scientist to learn about
cognitive load?
And, once we have a satisfactory answer to the questions above, how do we assess student
progress? Does this require multiple instructors with different backgrounds? Finally, how
5
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
do we scale this approach to larger groups of students, and how do we adapt it to online
learning?
Clearly, we do not yet have the answers to the above questions. In order to start finding the
answers, workshop participants worked in small groups to create new educational material
for topics of interest. This was a valuable exercise because it further helped sharpen the
questions in our minds. The small groups tested the new material on each other, allowing
for further in depth exchange of ideas.
OUTLOOK
Ubicomp education is complex. The field has a broad set of stakeholders: students,
educators, industry, government, as well as the general public. Each of the stakeholders
has multiple communities with different backgrounds, needs, and expectations. In five busy
and exciting days at Dagstuhl, we identified a number of issues to consider in ubicomp
education and training, and we identified a number of noteworthy approaches that are
currently used at institutions around the world. We also found that large unresolved issues
remain. One such issue is the very definition of what ubicomp is and what it is not. Another
is that we still do not quite know how best to address (and take advantage of) the
heterogeneity of our students. Furthermore, we do not yet know how best to align
educational approaches with the goals of various stakeholders. On the other hand, the
workshop also made it clear that there are a variety of well thought-out and successful
ubicomp educational approaches that work well for students. Perhaps most importantly,
our community embraces multidisciplinarity, and more broadly it embraces diversity: we
realize that the workforce of the future should be diverse in education, as well as along a
host of other dimensions. Such a diverse workforce is needed to support the needs of our
society.
Also, technological advances have provided educators with tools that can make a difference
in ubicomp education. We can train students to create devices, to collect measurements,
and to process data because the technology to do this has become accessible to most, both
in terms of cost and in terms of ease-of-use. And when students need training, they can
take advantage of a broad range of resources, from tutorials to full blown online programs,
from which they can find the help that they need.
6
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
So, what is next for ubicomp education? There are many exciting avenues to pursue. We
will wrap up this article with three such avenues—each of them connects teaching ever
advancing technologies with helping students understand the needs of users:
1. Doing good. The last few decades brought us unprecedented technological
advances that make it easier to implement our ubicomp ideas than it has ever been
before. But which ideas should we pursue? How to apply ethical lenses to evaluating
new ideas? Which ideas are the good ideas— those that will ultimately help us do
something beneficial in the world? We can think of these ideas as those that are
simultaneously useful, ethical, sustainable, and profitable. How do we teach our
students to find these good ideas? Furthermore, how do we teach them not to
pursue ideas that have harmful consequences?
2. AI and ubicomp. Advances in artificial intelligence (AI) have the potential to
transform the world.
5
But AI systems will not only be confined to situations where
they interact with individuals one-on-one (such as intelligent investment assistants
or robotic nurses), but will also be combined with pervasive computing technologies
and have world-wide implications. How can we prepare our students to create such
systems and to make sure that they implement good ideas, as in point one above?
3. Learning how to generate new knowledge. All human knowledge is advancing at
an increasingly rapid pace. We must help all students, and this includes students
preparing for careers related to ubicomp, to participate in the complex process of
generating new knowledge. For this, they will have to acquire many specific
tools—these are the tools we discussed in the section entitled “What?”. But, crucially,
they will also have to acquire the general tools of generating new knowledge:
understanding how to ask questions that have not been answered before, how to
propose hypotheses, how to test them, and how to turn what was learned into
products that impact us all.
Interested readers can find further information on the topic of ubicomp education in our
Dagstuhl report.
2
This includes insightful reflection statements by all of the participants, a
7
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
list of suggested reading, and a summary of possible active learning approaches for
ubicomp.
ACKNOWLEDGMENTS
The authors would like to acknowledge the support of Schloss Dagstuhl—Leibniz Center for
Informatics in organizing the seminar. The authors would also like to acknowledge all of the
participants of Dagstuhl seminar 19232 for contributing in myriad ways to the success of
the event. The work of Andrew Kun and Orit Shaer was supported in part by NSF grant
OISE-1658594. The work of Anne Roudaut was supported by the Bristol Institute For
Learning and Teaching.
REFERENCES
1. A. Girouard, A. L. Kun, A. Roudaut, and O. Shaer, “Pervasive computing education,” IEEE
Pervasive Comput., vol. 17, no. 4, pp. 9–12, Oct.–Dec. 2018.
2. A. Girouard, A. L. Kun, A. Roudat, and O. Shaer. “Ubiquitous computing education: Why,
what, and how (Dagstuhl Seminar 19232),” Dagstuhl Rep., vol. 9, no. 6, pp. 26–54, 2019.
3. A. McLeod, “Ubiquitous computing education,” IEEE Pervasive Comput., vol. 18, no. 3, pp.
59–62, Jul. 2019.
4. M. Weiser, “The computer for the 21st century,” IEEE Pervasive Comput., vol. 1, no. 1, pp.
19–25, Jan. 2002.
5. J. Brockman, ed., Possible Minds: Twenty-five Ways of Looking at AI. Baltimore, MD, USA:
Penguin Press, 2019.
AUTHORS
Andrew L. Kun is currently a Professor of electrical and computer engineering with the
University of New Hampshire, Durham, NH, USA. Contact him at andrew.kun@unh.edu .
8
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
Audrey Girouard is currently an Associate Professor with the School of Information
Technology, Carleton University, Ottawa, ON, Canada. Contact her at
audrey.girouard@carleton.ca .
Anne Roudaut is currently an Associate Professor with the Department of Computer
Science, University of Bristol, Bristol, U.K. Contact her at anne.roudaut@bristol.ac.uk .
Orit Shaer is currently the Class of 1966 Associate Professor of Computer Science and the
Co-Director of the Media Arts and Sciences Program, Wellesley College, Wellesley, MA, USA.
Contact her at oshaer@wellesley.edu.
9
PRE-PRINT FOR IEEE COMPUTER SOCIETY - PERVASIVE COMPUTING
