1
THE CHALLENGE OF INTEGRATING NEW TECHNOLOGY IN DESIGN EDUCATION
J. R. McCardle BSc PhD AMIEE
Department of Design & Technology, Loughborough University, UK
Keywords
Artificial intelligence, technology integration, research dissemination, industrial design, tertiary education
Abstract
This paper examines some of the many problems and issues associated with integrating new and developing
technologies into the education of future designers. As technology in general races ahead challenges arise
for both commercial designers and educators on how best to keep track and utilise the advances. The
challenge is particularly acute within tertiary education where the introduction of new cutting edge
technology is often encouraged. Although this is generally achieved through the feedback of research
activity, integrating new concepts at an appropriate level is a major task. Of particular concern is how
focussed areas of applied technology can be made part of the multidisciplinary scope of design education.
The paper describes the model used to introduce areas of Artificial Intelligence (AI) to undergraduate
industrial design students. The successful interaction of research and education within a UK higher
education establishment are discussed and project examples given. It is shown that, through selective tuition
of research topics and appropriate technical support, innovative design solutions can result. In addition, it
shows that by introducing leading edge and, in some cases, underdeveloped technology, specific key skills
of independent learning, communication and research methods can be encouraged. Furthermore, the paper
examines both the successes and failures of the process and provides conclusions relating to curriculum
development, effective learning, and assessment.
Introduction
If there is to be an increasing emphasis on the design of functional products within education it is essential
that design students gain a strong foundation in basic elements of technology. Specific areas of mechanics,
materials science and electrical/electronic engineering provide some of the information necessary to design
and construct a wide variety of working prototypes. The knowledge gained in these subjects allows students
to be more flexible in their approach to other areas of technology.
Industrial design education in the UK has been evolving steadily during the last two decades, with
developments paralleled in other countries. One of the issues that is being more systematically addressed is
the relationship between technology and designing. A report by Paul Ewing (1987) looked at this issue in
relation to the undergraduate and postgraduate education of industrial designers.
These skills that are so urgently needed, are being taught, in the author's view,
separately by two educational bodies. The Art Schools teach industrial design, and the
science-based academies, in some cases, teach engineering design. However there has
been little attempt to bring these two educational systems together. There is, though, a
light at the end of the tunnel, because a number of universities, polytechnics and colleges
of further education are beginning to realise the benefits of teaching design as a total
activity encompassing industrial and engineering design. (Ewing, 1987, p.2)
Ewing's report looked at the practice of twelve UK courses, four courses in the USA, one in Europe and one
in Japan that were at the time (1984) teaching some combination of industrial and engineering design. This
was followed by a more extensive survey of changing practices on industrial and product design courses by
Jeremy Myerson in 1991. This survey classified the technologies taught on UK industrial design courses
under ten headings: materials; processes; human factors; computing; workshop practice; manufacturing;
information management; engineering science; mechanical engineering; and electrical/electronic
engineering. A technological core was suggested (the first six of these categories), and recommendations
2
were made concerning its delivery, assessment, links with industry, staff development, information and
course titles. These were a valuable contribution, but the emerging nature of technological knowledge was
only alluded to e.g.
Many course leaders at both degree and HND level admitted they had problems in
defining a technological core of content. The following views were repeatedly expressed:
technology is fluid and ever-changing so there is no constant empirical body
of technical knowledge that can be defined and communicated;
technology will continue to change long after students have graduated so the
key strategy in the technological underpinning of course should be to imbue
students with a spirit of technical enquiry and give them skills to go researching
fresh technical information throughout their careers:
... (Myerson, 1991, p.28)
This paper seeks to contribute to this debate through the analysis of emerging AI technology as taught to
undergraduate industrial designers at Brunel University in the UK.
Many electronic consumer products are labelled ‘intelligent’ or ‘smart’ and the commercial implications of
using such technology are readily appreciated. However, the success of this technology depends on the way
in which it is designed into products for use in real world human situations. The way people react to
products is often a reflection on the ease with which they can interact with them, and applying AI can be
seen as a step towards the development of user-friendly products. Presently AI can be used to make very
limited human-like decisions and provide a form of interactive dialogue but, as a consequence, significant
questions arise concerning the user acceptability and perception of such interactive technology. At present,
important factors on how best to incorporate AI into products lack definition. Nevertheless its commercial
use makes an attractive proposition for students to tackle within industrial design. Potentially it provides
the opportunity to explore and experiment with state of the art technology and to develop innovative design
solutions.
Providing suitable information and guidance for such rapidly developing and commercially competitive
domains can, however, prove to be a major hurdle to educators. Due to the conflicting agendas of
Intellectual Property Rights (IPR) and research dissemination, the technology transfer from industry to
education (and indeed from education into industry) often lacks the dynamics required to support such an
initiative. However, in the area of industrial design higher education, institutions are often at an advantage
by being in a suitable position to exploit industrial collaborations as design solutions are often encouraged
to incorporate more ‘near market’ attributes.
Some research groups in the area of AI believe that exposing students to cutting edge technology, its
foundations, uses and development, can stimulate and yield innovative, technologically advanced design
solutions to real problems (McCardle 1998). In addition, it has been acknowledged by some research
institutions that education plays a major role in defining and advancing certain technologies adopted within
industry as students of today are potentially the end users and developers of tomorrow’s technology. It is
envisaged that accelerated developments can be achieved by introducing research-derived concepts at an
early stage in a student’s education (Ibid.).
Defining the Subject Area
The study of artificial intelligence requires attention to an eclectic body of knowledge. In a recent
benchmarking document for the UK Quality Assurance Agency (QAA), Aaron Sloman considered AI to be
a two-strand discipline of science and engineering, with science attempting to understand the mechanisms of
intelligence and engineering attempting to apply the findings in the design of useful machines, (Sloman,
2000).
3
Defining ‘intelligence’ is a philosophically problematic area and one that has, to date, not received an
acceptable consensus. Similarly, defining ‘artificial intelligence’ is equally troublesome but not only
through philosophical interpretation but also by what products, methods and techniques qualify for the title.
In an earlier work, Sloman defines AI as,
The general study of self modifying information-driven control systems,
both natural (biological) and artificial,
both actual (evolved of manufactured) and possible (including what might
have evolved but did not, or might be made at some future date) (Sloman, 1995)
Thus catering for both the science and engineering aspects of the discipline.
A less philosophical text book definition, and one perhaps more biased to an engineering approach, is given
by Elaine Rich & Kevin Knight,
Artificial Intelligence is the study of how to make computers do things which, at the
moment, people do better. (Rich & Knight, 1991, p.3)
There are many AI protagonists who will disagree with his statement simply because of its reference to
computers. The literal use of the term 'computer' implies that the advance of AI is as much a function of
Moore's Law (stating that computing processor power doubles every 18 months – although this itself is
thought to be conservative) as it is a true understanding of the underpinnings of intelligence.
When searching for a definition of AI, the then chief scientist at Apple, Lawrence Tesler introduced
Douglas Hofstadter to the dynamic nature of the field. His response to the question of defining AI has been
named 'Tesler's Theorem' by Hofstadter and simply states,
AI is whatever hasn’t been done yet (Hofstadter 2000, pp. 601)
There are various ways of interpreting this somewhat surprisingly succinct statement including claiming that
the field of AI widens, as more appropriate tasks become apparent. But additionally, if we take this
statement literally, it could also be interpreted that as specific tasks are accomplished then they cease to be
considered a legitimate area of AI. Although this provides a very ephemeral definition, historically this can
be seen to happen. Computers were once thought of as great electronic brains, and even the humble pocket
calculator, perhaps the most ubiquitous of all computers, was once thought of as state of the art in electronic
intelligence. Today we view such products with irreverence and they are universally accepted as common
useful tools. They are certainly not seen as representing any form of intelligence.
Considering the transient and fragmented nature of the subject it is generally considered that,
AI is better defined by indicating its range. (Oxford, 1997, p. 21)
To do so, however, yields a very subjective definition and one that necessitates the development of a
taxonomy.
Towards Developing an AI Taxonomy for Designers
For designers working primarily in the area of applications, what is perhaps more important than an airtight
definition of AI is the knowledge of the success or failures of particular existing techniques and the
feasibility of their use within certain products.
A common historical problem with most, if not all AI methodologies, has been the excessive claims made
about their capabilities, which often led to beliefs that the technology was a panacea to all computing
problems. However, over the last ten years, as the number of AI applications increased, the limitation of the
technology has become more evident. This has further resulted in increased scepticism from many industrial
4
sectors about their use. The reality is that AI methodologies are enabling technologies which, in a design
environment, can help to provide design improvements as well as complete solutions.
Table 1 illustrates some of the major areas of AI, which are presently being researched. Within these areas,
substantial and useful advances have been made, though it is acknowledged that most of these tasks are
considered non-trivial and remain underdeveloped.
Research Domain
Example Tasks
Communication
Natural language production, Syntax, Multi-agents,
Intelligent interfaces
Creativity and
Cognition
Cognitive modelling, creating works of art
Game Playing
Tactics (e.g. Chess), Virtual Reality (VR), Simulators
Information
Management
Knowledge based systems, Intelligent agents,
Data mining
Learning Systems
Machine learning, Behavioural modelling
Perception
Vision, (optical character recognition, text readers),
Spatial recognition, Hearing (voice recognition), smell
and taste recognition
Symbolic Logic
Methods of abstraction, Modelling techniques
Understanding
Comprehension, Natural language, Reasoning systems
Table 1.
Current AI Research Areas
Within each research domain, various techniques in modelling, abstraction and computation have evolved
and are often applied across the domains. The methods adopted can be dependent upon the application area,
for instance schemes which do not provide causal relationships or cannot successfully model system
functionality are rarely used for safety critical applications.
Table 2 divides AI approaches into what can be termed 'high' and 'low' levels. These levels refer to the
approaches to specific problems, the nature of the abstraction necessary and the complexity of the required
output. They do not refer to the complexity of the technique as every method, in the main, can be shown to
be non-trivial.
Low-level approaches tend to focus on ‘raw data’ applications and in that respect they are fundamental in
their approach. In general data can be pre-processed by a suitable algorithm and subjected to the technique
to reveal patterns and relationships, highlight anomalies, predict temporal sequences, filter, optimise and
learn patterns. The fundamental nature means that such techniques are portable from one problem to another
i.e. techniques that predict stock market trends can also be used to recognise patterns in speech analysis.
High-level approaches use techniques that use models of the environment to solve a specific problem. The
immediate difficulty with these approaches is setting the level of abstraction (often referred to as
granularity) which will adequately satisfy the constraints of the problem domain. As a result such tasks are
subjective and require a distinct and bespoke approach. The advantages are that holistic methods are
possible and the functionality of complete systems can be analysed to reveal causal relationships under
varying conditions.
5
The bespoke nature of high level approaches has, in general, made them specialist techniques requiring the
complex skills of programmers and engineers. In terms of applying the technology, few development tools
exist to enable the designer to construct and use model based techniques. This is, however, set to change as
researchers acknowledge that active technology transfer plays a substantial part in advancing the field (see
MONET 2000, for example).
In contrast, for the application of low-level approaches a plethora of commercially available software tools
exist to aid applications developers and educators. Consequently these techniques have provided a far more
accessible route to applying AI in product design.
High Level AI
Low Level AI
Task Domain
Typical Application
Task Domain
Typical Application
Reasoning Systems,
cognitive modelling,
Failure Modes Effects
Analysis (FMEA),
systems modelling,
diagnostics
Pattern recognition,
Prediction, Learning
schemes
Speech recognition,
image analysis, machine
health monitoring,
diagnostics, real time
control
Generic Techniques
Comments
Generic Techniques
Comments
Modelling, Abstraction
Can provide causal
relationships,
subjective, non-trivial,
time consuming,
computationally
intensive
Neural Networks,
Fuzzy Logic, Genetic
Algorithms
Portable, relatively short
development times, can
require large amounts of
data, often difficult to
analyse, non-formative
approaches are apparent
Table 2.
'High' and 'Low' Level Tasks and Techniques
IT and Smart Technology
Within education curricula, Information Technology has made a great impact expedited by the advent of the
World-Wide-Web and the consequent cultural information explosion. IT is a generic expression that can
include any form of technology (equipment and/or technique) used by people to collect, store, control and
communicate information. The field of AI encompasses all these areas and could therefore be considered a
part of the IT revolution.
'Smart Technology' is an unfortunate term that has been used in a somewhat cavalier manner, mainly due to
media hype and commercial exploitation. In the mid-eighties the development of shape memory alloys led
to the tag 'Smart Materials'. This was followed by 'Smart Cards' a simple memory device on a piece of
plastic the size of a credit card which contained user profiles to operate user dependent machinery such as
bank teller machines. Since then a whole manner of devices and products have been designed and marketed
bearing the label 'smart', or even more misleading, 'intelligent'. There are examples where IT has been
merged with common domestic products, for instance by allowing Internet access via an LCD screen in the
door of a microwave, or by installing a barcode reader and Internet access in a fridge, whereupon these
humble kitchen appliances are apparently rendered 'intelligent'.
A common misconception is that AI and Smart technology are one and the same thing. The understanding
and appreciation of AI as a philosophical science and an engineering discipline should, however, refute this.
The proposed model illustrated in Figure 1 aims to create a more usable definition for distinguishing AI and
Smart Technologies from an application based engineering perspective. If we consider Tesler’s theorem
then legitimate AI areas, in both high and low level techniques, are shown as future research goals.
Advances in AI research are disseminated through to applications research. Finally, proven techniques are
