ArticlePDF Available


This report is one of the outputs from the AIED theme of the Technology Enhanced Learning (TEL) research programme ( education/). This initial version of the text will be subject to revisions based upon feedback from readers. The Artificial Intelligence in Education (AIED) theme within the Personalisation strand of TEL is concerned with exploring the ways in which the work conducted under TEL within and across projects can contribute to the (inter)discipline of AIED.
What is AIED and why does Education need it?
A report for the UK's TLRP Technology Enhanced Learning - Artificial
Intelligence in Education Theme. May 2011.
Authors: Joshua Underwood and Rosemary Luckin, The London Knowledge Lab.
This report is one of the outputs from the AIED theme of the Technology Enhanced Learning
(TEL) research programme (
education/). This initial version of the text will be subject to revisions based upon feedback from
readers. The Artificial Intelligence in Education (AIED) theme within the Personalisation strand of
TEL is concerned with exploring the ways in which the work conducted under TEL within and
across projects can contribute to the (inter)discipline of AIED.
Education and AIED
AIED stands for Artificial Intelligence in Education, but is AIED research having an impact in
mainstream Education? Ten years ago Cumming and McDougal noted AIED could “scarcely
claim to be in Education” (2000, p.197). However, during the last decade AIED has made great
progress in moving out of labs into real world large-scale deployments that are having significant
impacts (see Table 1). Furthermore, tremendous technological advances, for example in mobile
systems and social networks, have been made and technologies, powerful enough to support AI
techniques such as user modelling and speech recognition are now pervasive throughout much of
society and our daily lives. The ubiquitous technology-rich settings that can enable widespread
uptake of AIED for more effective formal and informal learning are now in place, and while
developing robust solutions for AI problems is still challenging, such as speech recognition that
works well with continuous speech in noisy environments and with young learners, existing AI
technologies offer new opportunities for technology-enhanced learning.
However, other barriers to adoption and the everyday use of AIED systems remain. There are
many notable technological successes (see Table 1) and yet these technologies are not fully
exploited for educational purposes (Woolf, 2010) particularly within mainstream Education.
Cumming and McDougal observed that as more and “more of learners’ time will be spent with
technology (this) will bring to the fore in Education questions of how to design computer-based
learning resources that are effective, while retaining allegiance to Education’s rich
conceptualisations of learning.” They went on to claim that “(t)hese are precisely the issues on
which… …AIED claims insight, so Education should discover that it needs AIED!” (2000, p.203).
Ten years after the publication of Cumming and McDougal’s paper the relevance of AIED
research for today's teaching and learning challenges is not fully recognised in mainstream
Education. It may be that the AIED community has so far failed to successfully communicate with
educators and policy makers or provide the kinds of evidence and support required for wider
adoption. We believe AIED has the potential to make a much broader contribution to Education
than it currently is. AIED techniques can respond to today’s educational challenges by delivering
more flexible and inclusive, personalised, effective and engaging learning experiences throughout
lifetimes and across formal and informal settings. However, realising this potential requires a
greater effort to communicate the relevance of AIED research and to support its uptake within the
wider learning community and by other stakeholders in Education. It also requires that the AIED
community direct greater attention to understanding and communicating the manner in which
AIED technologies can be integrated into educational settings, for example through recognising
and developing a clear role for practitioners in relation to these technologies. Here, we describe
what AIED is, what it has to offer and why it is relevant now and for the future of Education.
What is AIED?
AIED is interdisciplinary research “at the frontiers of computer science, education and
psychology’. It promotes rigorous research and development of interactive and adaptive learning
environments for learners of all ages, across all domains” (International AIED Society, 2010).
The nature of the research conducted under the heading of AIED has developed and evolved
over the past 25 years and the community has become a broad church concerned with supporting
the learning and teaching process in situ and in real time. AIED research addresses learning
wherever it might occur, including in formal classroom settings as well as outside the classroom,
both to support formal education and for broader, lifelong learning. AIED researchers are paying
increasing attention to the affective and social as well as the intellectual aspects of learning with
very active research being conducted to investigate collaboration, meta-cognition, self-regulation,
motivation, and emotions, as well as the more traditional areas of scaffolding and intelligent
tutoring; and the new possibilities afforded by data mining techniques.
AIED research is driven by educational problems and is as much about a way of doing
research as about technology development. Theoretically grounded research is supported by
systematic empirical evaluation that informs further theory development. The AIED community is
actively exploring the ways in which learning and teaching can benefit from new and cutting edge
technology, particularly drawing on research in Artificial Intelligence (AI).
What is the AI in AIED for?
Artificial Intelligence is concerned with: creating computational models of human faculties (e.g.
speaking, learning, walking, and playing), enabling systems to replicate common-sense tasks (e.g.
understanding language, recognising visual scenes, summarising text) and otherwise reproducing
intelligent behaviour (Russell and Norvig, 1995). Intelligent behaviour may be usefully defined as
acting rationally; that is doing the right thing given the available information, taking actions
expected to maximise goal achievement (Russell and Norvig, 1995), which in the case of AIED
means acting to maximise learning both in the short and long term where learning is increasingly
understood in a very broad way including the development of personality, a sense of self-esteem
and self-efficacy among other forms of development. A key motivation for using AI techniques in
the development of Technology Enhanced Learning is to support the development of systems that
do the right things, or help teachers and learners to do the right things to maximise learning. This
involves understanding and modelling learners, teachers, effective pedagogies and learners’
contexts. AIED systems are adaptive, offering support dynamically by responding appropriately to
changing and incomplete information about learning objectives, learners, collaborators and
Why do Teaching, Learning and Education need AIED?
One of AIED’s scientific goals is to “…make computationally precise and explicit forms of
educational, psychological and social knowledge which are often left implicit” (Self, 1999, p. 1).
Such precise models of learning may become AIED’s greatest contribution to Education
(Cumming & McDougal, 2000). Modelling is a central concept in AIED and increasingly rich
models are being developed. Such models take into account characteristics of learners and
teachers, collaborative, social, affective and meta-cognitive aspects of learning, the settings
learning takes place in, and the learner’s context (e.g. Luckin, 2010). Baker (2000) identified three
main roles for AIED models:
1. Model as scientific tool. A model is used as a means for understanding or predicting
some aspect of an educational situation.
2. Model as component. A computational model is used as a system component enabling a
learning environment to respond adaptively to user or other input.
3. Model as basis for design. A model of an educational process, with its attendant theory,
guides the design of technology-enhanced learning.
In the intervening decade a fourth key role has emerged:
4. Open Models as prompts for learner and/or teacher reflection and action: Computational
models, usually of learner activity and knowledge, are made inspectable by and possibly
opened for learners and/or teachers to edit. Such open models can prompt users to reflect
on their learning and support meta-cognitive activity (Dimitrova, McCalla & Bull. 2007).
Models of these types and the adaptive systems that employ them have great utility for
Education research and practice both in terms of developing our understanding of educational
situations and learning and in delivering more efficient, personalised and contextualised support
for learning and teaching. Furthermore, deployed technology enhanced-learning environments can
be used to gather data efficiently and on a large scale, both to test and refine our understanding of
learning and teaching, and in order to provide evidence for the effect of these systems on learning.
Evaluation has been a consistent and increasingly important theme in AIED research (see
Underwood & Luckin, 2011 for an exploration of themes in AIED research). There is now
considerable work in evaluating AIED systems, particularly in well-defined domains, that are
applied on a global scale with hundreds of thousands of students in classroom or university
settings (see Table 1 for examples) (Dimitrova, 2010).
Table 1 Example ‘Mainstream’ Intelligent Learning Environments
For Learning Foreign Culture & Language
Tactical Language & Culture Training System (TLCTS)
Alelo’s Tactical Language and Culture Training System uses a virtual game-based
environment and interactive lessons to provide foreign language and culture training. TLCTS
employs AI techniques to process learners’ speech, engage in dialogue and evaluate
performance and has been used by more than 40,000 learners worldwide with independent
evaluations showing significant gains in learners’ knowledge of language and culture and
greater self-confidence in communicative ability (Johnson & Valente, 2009). See for more information.
For Learning Maths
Cognitive Tutors
Carnegie Learning’s Cognitive Tutors use AI techniques to provide learners of Maths with
individualized attention and tailored material based on continual assessments (Carnegie
Learning, Applying Cognitive Science to Education). Cognitive Tutors aim to act like human
tutors constantly monitoring learner actions and guiding learners towards correct solutions,
providing help on demand and in response to common mistakes and giving meaningful
feedback to students on their acquisition of skills (Carnegie Learning, The Cognitive Tutor
Successful Application of Cognitive Science). Cognitive Tutors are used in many schools in
the US and elsewhere and several evaluations of Cognitive Tutors have been conducted (see
Carnegie Learning, 2010). Evaluations have demonstrated that Cognitive Tutors can improve
problem solving and critical thinking skills (Koedinger, Anderson, Hadley & Mark, 1997),
improve performance on exams (Sarkis, 2004), improve student attitudes to mathematics
(Morgan & Ritter, 2002), and show strong results for disadvantaged populations (Sarkis,
2004). See for more information.
Wayang Outpost
Wayang Outpost is an intelligent tutoring system that helps learners prepare for maths tests
and helps teachers in their assessment of students' strengths. Wayang can provide interactive
hints leading to the solution for a problem. As the student progresses through problems the
system adjusts instruction using individualized strategies that are effective for each student.
An evaluation of Wayang (Beal, Walles, Arroyo, Woolf, 2007) shows significant
improvements on pre to post-tests and suggests the greatest benefits are for weaker students
and those who make most use of the multimedia help features. For more information and to
register to try the system out see
ActiveMath is an adaptive learning environment for Mathematics that applies AI techniques to
automatically assemble individualised courses. ActiveMath can generate courses adapted to
the learner’s curriculum, language and field of study, as well as to her cognitive and
educational needs and preferences such as learning goals, preferred style of presentation, goal-
competencies, and mastery-level (Melis & Siekmann, 2004). ActiveMath includes interactive
exercises that can provide feedback and hints of different kinds in response to learner input.
The ActiveMath system has been used and evaluated in classrooms and universities in various
European countries for several years (see A
Europe-wide formative and summative evaluation investigated usability and learners’ opinions
of automatically generated courses; results indicated that learners appreciated the generated
courses, felt these were personalized and that the generated courses helped learners to find
their own way of learning (Ullrich & Melis, 2010). For more information about the
ActiveMath system, research and access to a demonstration version see
For Learning Physics
Andes Physics Tutors
Andes is an intelligent homework helper for Physics. Students enter steps in solving a
problem, such as drawing vectors, drawing coordinate systems, defining variables and writing
equations and Andes provides feedback after each step (VanLehn et al, 2005). Andes
encourages learners to use good problem solving strategies, provides immediate feedback on
learner input and offers different kinds of instructional assistance depending on the kinds of
error learners make. Andes has been used successfully since 2000 in the US Naval Academy
and is in use elsewhere at college and high school level (see for
more information). Evaluations in real classrooms over five years show that Andes is
significantly more effective than doing pencil-and-paper homework and at a low cost, with
students spending no extra time doing homework, and with no need for teachers to revise their
classes in order to obtain these benefits (VanLehn et al, 2005). The Andes Physics Tutor is in
use on an Open Free Physics course provided through the Open Learning Initiative.
For Learning Programming and Database skills
SQLTutor, Database Place & ASPIRE
SQLTutor provides adaptive individualized instruction that helps learners’ master key
concepts in database courses using student and pedagogical models. SQLTutor has been in
large-scale use with several thousand users (Mitrovic et al., 2006), evaluated on numerous
occasions and refined for more than a decade (see SQLTutor Evaluations). Evaluations of
SQLTutor have demonstrated the need for feedback to be personalized to individual students’
needs (Martin & Mitrovic, 2006) the value of both negative and positive feedback, as opposed
to only negative feedback, with students receiving both forms of feedback requiring
significantly less time to solve the same number of problems, in fewer attempts and learning
the same number of concepts as students in the control group (Barrow, Mitrovic, Ohlsson, &
Grimley, 2008). SQLTutor is one of a number of constraint-based Intelligent Tutoring
Systems (ITSs) produced by the Intelligent Computer Tutoring Group (ICTG) at University of
Canterbury (New Zealand). These ITSs have proven effective not only in controlled studies
but also in real classrooms, and some of them have been commercialized (Mitrovic et al,
2009). SQLTutor and other adaptive tutors for database skills are available through Addison-
Wesley's Database Place. ICTG are also working towards making it easier for teachers and
domain experts to develop ITSs. ASPIRE (Authoring Software Platform for Intelligent
Resources in Education) assists users in developing and delivering online constraint-based
tutors and is freely available to all New Zealand Government-owned Tertiary Institutions. For
more information about Intelligent Tutors developed by ICTG see
ELM-ART: Episodic Learner Model - The Adaptive Remote Tutor
ELM-ART is an intelligent interactive system that supports learning to programme in LISP.
“ELM-ART provides all learning material online in the form of an adaptive interactive
textbook… …ELM-ART provides adaptive navigation support, course sequencing,
individualized diagnosis of student solutions, and example-based problem-solving support.”
(Weber & Brusilovsky, 2001, p.351). Provision of the system online was found to greatly
contribute to flexibility and efficiency of learning with students accessing the system from
both home and university locations, with many students completing the course in very short
periods of time and achieving very good results in the final programming task (Weber &
Brusilovsky, 2001). One AIED approach employed in ELM-ART is adaptive link annotation.
Adaptive annotations augment hyperlinks with personalised hints that can help guide learners
to the most personally appropriate learning content at any given moment. Adaptive annotation
has been adopted by many systems and “(e)mpirical studies of adaptive annotation in the
educational context have demonstrated that it can help students to acquire knowledge faster,
improve learning outcomes… (and) significantly increase student motivation to work with
non-mandatory educational content” (Brusilovsky, Sosnovsky & Yudelson, 2006, p.51). ELM-
ART has been used over many years by hundreds of students to support delivery of a
university course. You can try out ELM-ART at
KnowledgeSea II
Knowledge Sea II is a mixed corpus C programming resource that bridges the gap between
closed corpus materials in the form of lecture notes and open-corpus materials in the form of
links to online resources for C programming. Knowledge Sea II helps users navigate from
lectures to relevant online tutorials by providing links to online material related to search
keywords. Search is adapted to the user by taking into account both the past interactions of the
individual user and the user’s group (other learners). KnowledgeSea prompts learners to
access material related to the user’s search by providing traffic and annotation based social
navigation support. Social navigation support is realised by marking links to material with
icons and colour codes that indicate the amount of traffic (time spent reading the linked
material by other learners) and positive and negative individual and group annotations of the
linked material (Brusilovsky, Farzan, & Ahn, 2006). Evaluations of KnowledgeSea II show
that pages automatically predicted as important for a learner were actually rated as important
by students and that the adaptive link annotations successfully influenced learner behaviour,
with learners preferentially accessing more highly ranked pages and those with link
annotations that indicate higher traffic (Brusilovsky, Farzan, & Ahn, 2006). For more about
KnowledgeSea see You can
register to try the system at
AIED, Productivity, Personalisation, Inclusion and Flexibility of Learning
“In looking to TEL for improvements in productivity we need to look for ways of:
- Improving the quality of teaching in order to improve the quality of learner achievement
against the learner’s time
- Increasing the number of learners achieving quality outcomes against teacher time
- Reducing the amount of teacher or learner time needed for learner achievement”
(TLRP-TEL Programme, 2010)
Personalisation improves the use of learner time by enabling learners to work at their own pace,
receive targeted feedback, and be supported in their learning without relying on teacher presence
(TLRP-TEL Programme, 2010). AIED systems can deliver such learner control but they may also
employ models of learners and pedagogic strategies to seek to engage learners and push them
when necessary. Enabling systems to deliver such personalisation is a key driver of much AIED
and User Modelling research. Throughout the last decade various intelligent techniques that
contribute to the personalisation of learning have been developed and empirically shown to be
effective (Dimitrova, 2010); these include:
modelling the learner’s cognitive states to provide individualised learning (VanLehn,
using tutoring dialogues, even with shallow natural language processing, to deepen
learning experiences (Litman, 2009);
using open learner models to promote reflection and self-awareness (Bull & Kay, 2007;
Dimitrova, McCalla & Bull, 2007; Mitrovic & Martin, 2007);
adopting meta-cognitive scaffolding to increase learner motivation and engagement (du
Boulay, Rebolledo Mendez, Luckin & Martinez Miron, 2007; Harris, Bonnett, Luckin,
Yuill & Avramides, 2009).
Such AIED systems can now be deployed online and to personal and portable devices within
and beyond formal educational settings and consequently can also contribute to both the flexibility
of learning and to greater inclusion.
Flexibility is a way of improving the use of both learner and teacher time, face-to-face teaching
can be replaced by online teaching, individual learning and group work (TLRP-TEL Programme,
2010). Many AIED systems are now web-based (see Table 1) and AIED researchers are exploring
the use of mobile devices to deliver adaptive materials for more flexible anytime anywhere
learning. Social and collaborative aspects of learning are increasingly important themes in AIED
research and systems that monitor group work and provide effective intelligent support for
collaboration, both at a distance and face-to-face, are being developed (e.g. Upton & Kay, 2009).
Inclusion is a way of increasing the number of learners attaining a particular level: attracting
The TLRP-TEL website
disaffected learners through more engaging forms of learning; providing additional help for
learners with special needs; and motivating learners who cannot attend school (TLRP-TEL
Programme, 2010). Motivation and Affect have been major themes in AIED research throughout
the last decade (see Underwood &Luckin, 2011). There is substantial on-going AIED research into
the use of games to deliver more engaging learning experiences (Johnson, 2010) and much AIED
research employing novel user interfaces (e.g. natural language, speech and gesture recognition,
eye-tracking and physiological and other sensors), which offer opportunities to engage learners
with widely differing needs. Some systems (e.g. CognitiveTutors) have demonstrated strong
results for disadvantaged populations (Sarkis, 2004). AIED researchers are also concerned to
develop methods that enable learners, including children (Good & Robertson, 2006) and those
with specific needs (Porayska-Pomsta, Bernardini & Rajendran, 2009), to participate in the design
of systems that meet these users’ particular requirements. Other current and emerging themes in
AIED research (e.g. adaptive support for inquiry learning, exploratory learning, lifelong learning
and learning in ill-defined domains see Underwood & Luckin, 2011) will also contribute to
greater inclusion and flexibility of learning in the near future.
Grand Challenges for future AIED
AIED research is driven by the need to respond to educational challenges and to develop and
exploit new technologies in pursuit of solutions to these challenges. Several Grand Challenges for
future technology enhanced learning have been identified over the last decade. In the UK Grand
Challenges for Computing include Learning for Life (Taylor et al, 2008). This challenge identifies
requirements for future systems: to model learners and support dynamic evaluation and
assessment, to develop facilities that support learning outside formal educational settings over a
learner’s lifetime, to encourage and support creativity and problem solving and to ensure that
learning for life is a viable option for all through enhanced accessibility and inclusion. In the US
Grand Research Challenges in Information Systems identifies the need to “provide a teacher for
every learner” asserting that in the future “individually tailored learner-centred tutoring will enable
people to more fully realize their potential”(Computing Research Association, 2003, pp. 2). More
recently the National Academy of Engineering (2010) identifies the need to “advance personalized
learning” as a Grand Engineering Challenge recognising that new and emerging technologies offer
the potential for instruction to be “individualized based on learning styles, speeds, and interests to
make learning more reliable”. These are amongst the key challenges that AIED responds to.
Following on from the identification of such challenges a recent NSF funded report, GROE: A
Roadmap for Educational Technology (Woolf, 2010), considers the role educational technology
needs to play in advancing education over the next 30 years. The report draws on findings from
several workshops, led by a large team of international experts. These workshops focussed on
identifying promising technologies with which to address seven key educational challenges
described as: personalizing education, assessing student learning, supporting social learning,
diminishing boundaries, developing alternative teaching strategies, enhancing the role of
stakeholders, and addressing policy changes” (Woolf, 2010 pp.6). The report goes on to identify
seven particularly promising areas of technology research and development that offer opportunities
to satisfy the educational challenges identified: user modelling, mobile tools, networking tools,
serious games, intelligent environments, educational data mining, and rich interfaces. User
Modelling and Intelligent Environments are central to AIED research and the other technologies
identified are major themes in AIED research over the last decade (see Underwood & Luckin,
2011). Visions for next generation technology-enhanced learning integrate expected advances in
these areas of research to deliver solutions to the educational challenges identified earlier.
What will next generation AIED learning environments be like?
In this section we summarize visions of next generation learning environments, as described in the
GROE report (Woolf, 2010), in order to highlight the expected role of AIED research.
Intelligent teaching and learning environments, incorporating sophisticated user models, will
provide flexible and adaptive assistance, personalised to individual learners needs. Such, assistance
will be domain independent and will include support for “soft skills, such as creativity, critical
thinking, communication, collaboration, information literacy, and self-direction, and will be open-
ended and exploratory in nature, allowing learners to question and enhance their understanding
about areas of knowledge in which they are motivated to learn” (Woolf, 2010 p58). Personalised
support and feedback will be available to learners across subjects and across formal and informal
settings and throughout their lifetimes. Open user models will prompt learners to reflect on their
learning and how they learn. Such systems, accessible ubiquitously through mobile and distributed
rich interfaces, will improve flexibility and inclusion and help dissolve the boundaries between
sites of learning and connect learning across subjects (Woolf, 2010).
Future intelligent teaching and learning environments may be self-learning, using machine-
learning techniques to learn about students and improve their own performance by evaluating how
they are used and associated with learning outcomes. However, such environments will not aim to
replace teachers but rather work with teachers both being informed by teachers’ input and
informing teachers’ decisions and actions. These environments will take into account the needs
and interests of learners and will employ novel interfaces and techniques from games to deliver
highly engaging and accessible learning experiences. Next generation learning environments will
augment the real world with interactive representations that support learning through interaction
with tangible physical world and mixed reality interfaces while making perceptible, phenomena
that are too large, too small, too quick and too slow to observe in the real environment (Woolf,
2010). Simulations and micro-worlds will support learning through exploration and will provide
appropriate and personalised feedback. These environments will enable learners to move
seamlessly between real and virtual worlds and will span formal and informal activities and will
better connect learning across these worlds (Woolf, 2010).
Intelligent teaching and learning environments will also use networking tools to better support
social and collaborative learning, introducing suitable collaborators, guiding learners towards
effective collaboration and helping teachers to monitor and support group work. Such networking
tools “will facilitate individuals to learn within communities, communities to construct knowledge,
and communities to learn from one another” (Woolf, 2010 p 51.). In these communities, learner
roles will be more fluid, with teachers often acting as facilitators and opportunities for learners to
participate as producers and teachers as appropriate to their knowledge and interests. Ubiquitous
access to such communities will contribute to diminishing boundaries between formal and
informal learning (Woolf, 2010).
Educational data-mining techniques will be used to discover patterns in the vast amounts of
data that become available from such integrated intelligent learning environments and will support
identification of success factors and problems (Woolf, 2010). These analyses will inform and
guide stakeholders, including teachers, parents and policy makers. Teachers will be supported with
easily accessed, more accurate and timely information and analysis of individual and group
learning. This information will support teachers in making strategic decisions and providing
appropriate guidance, and in the continuous assessment of learning. Appropriate information may
also be shared with parents enabling them to provide additional help and motivation (Woolf,
2010). Educational data-mining, an emerging discipline concerned with developing methods to
explore data from educational settings and better understand students and the settings they learn
, will reveal new challenges and opportunities for the AIED community. The rich data captured
by these systems will provide researchers with new opportunities to evaluate models of learning
and develop theory (Woolf, 2010).
AIED research offers some solutions to today’s educational challenges and has the potential to
deliver more flexible and inclusive, personalised, effective and engaging learning experiences
throughout lifetimes and across formal and informal settings. Over the last decade AIED research
has made substantial progress in demonstrating this potential by moving out of labs into large-
scale deployments that are having impacts in real-world settings as illustrated by the systems
described in Table 1. The technological infrastructure required to deliver AIED learning
experiences is increasingly ubiquitous. Current research in AIED aims to develop more flexible
systems that will increase access to effective, personalised and engaging, anytime, anywhere
learning throughout lifetimes across the full range of knowledge domains and skills and employing
varied pedagogic approaches. Realising this potential will certainly involve overcoming technical
obstacles, but mainstreaming AIED into Education will also require much more. It will require the
successful communication of the value of AIED research and systems. In particular, the role that
AIED systems can play within the broader educational settings of their use and with respect to the
other resources available to learners, such as teachers, peers and the physical features of the
environment, must be more clearly explained. Cumming and McDougal (2000) suggested one key
reason AIED had not been taken seriously in Education ten years ago was the use of insufficiently
rich models of learning. In the intervening decade one of the main focuses of AIED has been to
For information on Educational Data Mining see
develop much richer models of learning, learners, teachers and, to a lesser extent learners’
contexts. However, AIED research has typically been published in specialist journals and
conference proceedings, which have not been sufficiently visible beyond the community and are
only now becoming more visible in educational research resources (e.g. ERIC
). The AIED
community needs to better explain to Education the nature of the models used and the value of
these and the systems developed using them.
Even within AIED it is difficult and time consuming to keep track of, integrate and synthesise
work happening in all the specialist areas, let alone combine components into working systems.
The AIED and ITS conferences do provideopportunities for the cross-fertilization of approaches,
techniques and ideas from the many areas that make up this interdisciplinary research field,
including: agent technologies, artificial intelligence, computer science, cognitive and learning
sciences, education, educational technologies, game design, psychology, philosophy, sociology,
anthropology, linguistics, and the many domain-specific areas for which AIED systems have been
designed, deployed and evaluated” (AIED Conference, 2011
). However, this increasing multi-
disciplinarity is both a great strength of AIED and a massive challenge. Communication across
disciplines is notoriously difficult and integration of approaches and conceptual frameworks even
more so (Conole et al, 2010). AIED needs to communicate successfully both within the field and
beyond, particularly with mainstream Education. The need to support research that harnesses and
integrates knowledge across multiple disciplines to create a common groundwork of
conceptualization, experimentation and explanation that anchor new lines of thinking and inquiry
towards a deeper understanding of learning”
is recognised in the NSF Science of Learning
Centers program, with The Pittsburgh Science of Learning Center (PSLC)
aiming to enhance
scientific understanding of robust learning in real educational settings by bringing together basic
and applied research and supporting field-based experimentation with AIED systems. New
technologies offer opportunities to support such communication and interdisciplinary research (e.g.
see PSLC Robust Learning Theoretical Framework Wiki
and work in the TLRP Technology
Enhanced Research theme
) The need for such resources is recognised within the AIED
community, Woolf (2009) suggests “we need: cadres of bibliographies, suites of project
inventories, component exchange communities and global networks of test beds for intelligent
learning environments”. These resources would be helpful both within the discipline and beyond,
facilitating access to specialist AIED research and easier, faster development of learning
environments that incorporate intelligent components. However, substantial work is required to
specify requirements for such resources and this in itself will require collaboration between AIED
researchers, the wider Education community and other stakeholders.
Many thanks to our Advisory Committee: Vania Dimitrova, Judy Kay, Paul Brna, Kaska
Porayska-Pomsta, Chee Kit Looi, for their invaluable input to this report.
Education Resources Information Center - ERIC is progressively
indexing the International Journal of Artificial Intelligence in Education and at 31/01/2011 listed a
total of 11 records, this compares to 1120 records from Computers & Education on the same date.
About the AIED 2011 conference -
United States National Science Foundation - Science of Learning Centers -
LearnLab Pittsburgh Science of Learning Center -
PSLC Robust Learning Theoretical Framework Wiki -
UK Teaching & Learning Research Programme – Technology Enhanced Research Strand -
Baker, M. (2000). The roles of models in Artificial Intelligence and Education research: a
prospective view. International Journal of Artificial Intelligence in Education, 11, pp. 122-143.
Barrow, D., Mitrovic, A., Ohlsson, S., Grimley, M. (2008). Assessing the impact of positive
feedback in constraint-based ITSs. B. Woolf et al. (eds) Proceedings of the 9th Int. Conf. ITS
2008, LCNS 5091, Springer-Verlag, pp. 250-259.
Beal, C. R., Walles, R., Arroyo, I., and Woolf, B. P. (2007). On-line tutoring for math achievement
testing: A controlled evaluation. Journal of Interactive Online Learning, 6 (1), pp. 43-55.
Bull, S. & Kay, J. (2007). Student Models that Invite the Learner In: The SMILI Open Learner
Modelling Framework, International Journal of Artificial Intelligence in Education, 17(2), pp.
Brusilovsky, P., Farzan, R., and Ahn, J-w. (2006). Layered Evaluation of Adaptive Search. In: R.
W. White, G. Muresan and G. Marchionini (eds.) Proceedings of Workshop on Evaluating
Exploratory Search Systems at SIGIR 2006, Seattle, USA, August 10, 2006, pp. 11-13.
Available online at
Brusilovsky, P., Sosnovsky, S., and Yudelson, M. (2006) Addictive links: The motivational value
of adaptive link annotation in educational hypermedia. In: V. Wade, H. Ashman and B. Smyth
(eds.) Proceedings of 4th International Conference on Adaptive Hypermedia and Adaptive Web-
Based Systems (AH'2006), Dublin, Ireland, June 21-23, 2006, Springer Verlag, pp. 51-60.
Carnegie Learning, Applying Cognitive Science to Education. Available at
Carnegie Learning, Successful Application of Cognitive Science. Available at
Carnegie Learning (2010). Cognitive Tutor Effectiveness – Executive Summary. Available at
Computing Research Association (2003). Grand Research Challenges in Information Systems
Available at
Conole, G., Scanlon, E., Mundin, P., and Farrow, R. (2010). Interdisciplinary research - Findings
from the Technology Enhanced Learning Research Programme. TLRP, UK. Available
19/01/2010 at
Cumming, G., and McDougal, A. (2000). Mainstreaming AIED into Education? International
Journal of Artificial Intelligence in Education, 11, pp. 197-207.
Dimitrova, V (2010). ImREAL project:
Dimitrova, V., McCalla, G., and Bull, S. (2007). Open Learner Models: Future Research
Directions - Special Issue of the IJAIED (Part 2). International Journal of Artificial Intelligence
in Education, 17, 3, pp. 217-226.
du Boulay, B., Rebolledo-Mendez, G., Luckin, R., Martinez-Miron, E.A., and Harris, A. (2007).
Motivationally Intelligent Systems: Diagnosis and feedback. In R. Luckin, K. R. Koedinger & J.
Greer (Eds.), AIED 2007, IOS Press. Frontiers in Artificial Intelligence and Applications, 158,
pp. 563-565.
Good, J., & Robertson, J. (2006). CARSS: A Framework for Learner-Centred Design with
Children. International Journal of Artificial Intelligence in Education, 16 (4) pp. 381-413.
Harris, A., Bonnett, V., Luckin, R., Yuill, N., Avramides, K. (2009). Scaffolding effective help-
seeking behaviour in mastery and performance oriented learners. In V. Dimitrova, R.
Mizoguchi, B. du Boulay, A. C. Graesser (Eds.), AIED 2009, IOS Press. Frontiers in Artificial
Intelligence and Applications, 200, pp. 425-432.
International AIED Society (2010). About the society. At checked 31/01/11
Johnson, W.L. (2010). Serious Use of a Serious Game for Language Learning. International
Journal of Artificial Intelligence in Education, 20 (2) pp. 175-195.
Johnson, W.L., Valente. A. (2009) Tactical Language and Culture Training Systems: Using AI to
Teach Foreign Languages and Cultures. AI Magazine, 30 (2) pp. 72-83. Available online at
Koedinger, K. R., Anderson, J. R., Hadley, W. H., and Mark, M. A. (1997). Intelligent tutoring
goes to school in the big city. International Journal of Artificial Intelligence in Education, 8,
Litman, D. (2009). The Evolution of Natural Language Processing in AIED: Successes and
Challenges. Panel on the Evolution of AIED @ AIED09, July, 2009, Brighton, UK
Luckin, R. (2010). Re-designing Learning Contexts: Technology-rich, Learner-centred Ecologies.
Routledge, London.
Martin, B., Mitrovic, A. The effect of adapting feedback generality in ITS. In V. Wade, H.
Ashman, and B. Smyth (Eds.): AH 2006, LNCS 4018, pp. 192-202, 2006.
Melis, E., & Siekmann, J.H. (2004). ActiveMath: An Intelligent Tutoring System for Mathematics.
In L. Rutkowski, J.H. Siekmann, R. Tadeusiewicz, L.A. Zadeh (Eds.): Artificial Intelligence
and Soft Computing - ICAISC 2004, 7th International Conference, Zakopane, Poland, June 7-
11, 2004, Proceedings. LNCS 3070, pp. 91-101, Springer.
Mitrovic, A. & Martin, B. (2007). Evaluating the Effect of Open Student Models on Self-
Assessment. International Journal of Artificial Intelligence in Education, 17(2), pp.121-144.
Mitrovic, A. & the ICTG team, (2006). Large-Scale Deployment of three intelligent web-based
database tutors. Journal of Computing and Information Technology, 14 (4), pp. 275-281.
Mitrovic, A., Martin, B. Suraweera, P., Zakharov, K., Milik, N., Holland, J., McGuigan, N. (2009)
ASPIRE: An Authoring System and Deployment Environment for Constraint-Based Tutors.
International Journal of Artificial Intelligence in Education, 19,(2), pp. 155-188.
Morgan, P., & Ritter, S. (2002). An experimental study of the effects of Cognitive Tutor Algebra I
on student knowledge and attitude. Available at
National Academy of Engineering (2010). Grand Challenges for Engineering – Advance
Personalise Learning. Available at
Porayska-Pomsta, K. Bernardini S., Rajendran, G. (2009) Embodiment as a means for Scaffolding
Young Children's Social Skill Acquisition. In Proceedings of the Workshop on Children and
Embodied Interaction: Seeking Common Ground, 8th International Conference on Interaction
Design and Children (IDC'09), Como, Italy, 2009
Russell, S., & Norvig, P. (1995). Artificial Intelligence: A Modern Approach. Prentice-Hall,
Saddle River, NJ.
Sarkis, H. (2004). Cognitive Tutor Algebra 1 Program Evaluation: MiamiDade County Public
Schools. Lighthouse Point, FL: The Reliability Group.
Self, J. (1999). The defining characteristics of intelligent tutoring systems research: ITSs care,
precisely. International Journal of Artificial Intelligence in Education, 10, pp. 350-364.
Taylor, J., Greenwood, Wood, W., Rae, J., Rico, M. (2008) A Grand Challenge for Computing:
Learning for Life. Available at
TLRP-TEL Programme (2010). Productivity: Achieving higher quality and more effective learning
in affordable and acceptable ways. Available at
Ullrich, C., & Melis, E. (2010). Complex Course Generation Adapted to Pedagogical Scenarios
and its Evaluation. Educational Technology & Society, 13 (2), pp. 102115.
Underwood, & Luckin, R., (2011). Themes and trends in AIED research, 2000 to 2010. A report
for the TLRP. Available from
Upton, K., & Kay, J. (2009). Narcissus: Group and Individual Models to Support Small Group
Work. User Modeling, Adaption, and Personalisation. LNCS, Vol. 5535/2009, pp. 54-65.
VanLehn, K. (2006). The Behavior of Tutoring Systems. International Journal of Artificial
Intelligence in Education, 16(3), pp. 227-265.
VanLehn, K., Lynch, C., Schultz, K., Shapiro, J. A., Shelby, R. H., Taylor, L., Treacy, D.,
Weinstein, A., and Wintersgill, M. (2005). The Andes physics tutoring system: Lessons learned.
International Journal of Artificial Intelligence and Education, 15(3), pp. 147-204.
Weber, G., & Brusilovsky, P. (2001) ELM-ART: An Adaptive Versatile System for Web-based
Instruction. International Journal of Artificial Intelligence in Education, 12, pp. 351-384.
Woolf, B. (2010). A Roadmap for Education Technology. GROE Available online at
20Final%20Report.pdf checked 31/01/11
Woolf, B. (2009). AIED Grand Challenges. Panel on the Evolution of AIED @ AIED09, July,
2009, Brighton, UK
... AIEd augments teachers' intelligence by providing them with sets of predictions and recommendations to maximize learning (personality growth, development in the sense of self-efficacy and selfesteem) (Underwood & Luckin, 2011). AIEd pivots around models which are designed by traits of students, teachers, affective, metacognitive, collaborative factors of learning, the learning environment and the context of learners (Luckin, 2010). ...
Full-text available
Emotions play a vital role in self-regulated learning (SRL)processes and drastically influence cognitive functioning. Along with creating individualized, engaging, flexible and inclusive learning environments, Artificial intelligence (AI) learning systems, especially intelligent tutoring systems (ITSs) have the potential to sense affective states of a learner and respond to them to maintain learning flow. This paper discusses the concept of AI in education (AIEd) followed by the explanation of the role of emotions in SRL. Highlighting theoretical and technical aspects, it provides a discussion of ITS with an example of its benefits in learning.It also overviews affect detection and responding in affect sensitive ITSs. Before concluding, the paper highlights some limitations of the AI learning systems to detect affective states and achieve maximum student learning outcomes.
... As regards classroom, Underwood and Luckin (2011) is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be created to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans and improve themselves" (Russel & Norvig, 2010, p. 17 Baker and Smith (2019, p. 10) define AI as "computers which perform cognitive tasks, usually associated with human minds, particularly learning and problem-solving". ...
Full-text available
This paper explores the field of Artificial Intelligence applied to Education, focusing on the English Language Teaching. It outlines concepts and uses of Artificial Intelligence, and appraises the functionalities of adaptive tools, bringing evaluative feedback on their use by American school teachers, and highlighting the importance of additional research on the matter. It was observed that the tools are valid media options to complement teaching, especially concerning adaptive learning. They offer students more inclusive opportunities: they maximize learning by tailoring instruction to address students ‘needs, and helping students become more responsible for their own schooling. As for teachers, their testimonials highlight the benefits of dedicating more class time to the students’ most pressing weaker areas. Drawbacks might include the need to provide teachers with autonomy to override recommendations so as to help them find other ways to teach a skill that seems to be more effective for a specific student.
... AIEd augments teachers' intelligence by providing them with sets of predictions and recommendations to maximize learning (personality growth, development in the sense of self-efficacy and selfesteem) (Underwood & Luckin, 2011). AIEd pivots around models which are designed by traits of students, teachers, affective, metacognitive, collaborative factors of learning, the learning environment and the context of learners (Luckin, 2010 AIEd would create optimal student-centered smart learning environments for allowing each student's learning personalized and tailored according to their learning preference (Seldon & Abidoye, 2018)and the development of 21 st -century skills. ...
Full-text available
Artificial intelligence (AI) learning systems have the potential to help learners cope with the demands of future work requirements. Learning with AI systems can usher students to develop 21st-century skills more effectively by providing individualized, engaging, flexible and inclusive learning environments. AI in education (AIEd)has been advantageous in maximizing student learning outcomes and can prepare students to thrive and contribute to the growing knowledge society and the future of automation. This paper provides a discussion of the 21 st-century skills and some flaws in the current education system to help students develop advanced skills. It also briefly elaborates the concept of AIEd concerning the development of 21 st-century competencies. Highlighting the existing application of AI learning tools and their potential, the paper explains their advantages in helping students develop the skills. Before concluding, it discusses some limitations of the learning systems.
... This involves understanding and modeling learners, teachers, effective pedagogies and learners' contexts. (Underwood and Luckin, 2011). Artificial Intelligence is concerned with: creating computational models of human faculties (e.g. ...
Purpose The last few years have seen a surge of interest in artificial intelligence (AI). The purpose of this paper is to capture a snapshot of perceptions of the potential impact of AI on academic libraries and to reflect on its implications for library work. Design/methodology/approach The data for the study were interviews with 33 library directors, library commentators and experts in education and publishing. Findings Interviewees identified impacts of AI on search and resource discovery, on scholarly publishing and on learning. Challenges included libraries being left outside the focus of development, ethical concerns, intelligibility of decisions and data quality. Some threat to jobs was perceived. A number of potential roles for academic libraries were identified such as data acquisition and curation, AI tool acquisition and infrastructure building, aiding user navigation and data literacy. Originality/value This is one of the first papers to examine current expectations around the impact of AI on academic libraries. The authors propose the paradigm of the intelligent library to capture the potential impact of AI for libraries.
The complexity of the current educational context, that is, the fragmentation of knowledge and the diversity which characterizes the students in our schools requires new competences: the learner should develop skills in building significant network, the teacher should set multiple dispositifs that are connected to support learning. The personalization of the processes is certainly an answer, that is, the chance to develop different paths in which diversity is respected, but in which also the autonomy of the learners is fostered. For sure the problem is the sustainability of such path. Can adaptivity be a strategy in this direction? If yes, how and thank to what modalities?
The complexity of the current educational context, that is, the fragmentation of knowledge and the diversity which characterizes the students in our schools requires new competences. One of the answer is certainly to think about personalized path. The personalization of paths is a hard task for the teacher and, thus, the efficacy and also the sustainability of this choice need to be examined. For what concerns sustainability a solution has been provided by the interaction between the educational world and the world of knowledge engineering. The focus on the user, on intelligent systems for personalization, the adaptive and responsive design are proposals that were born with a different objective, but that have opened new perspectives also in the educational context. Besides, such processes have linked different research fields, the one of education and the one of knowledge engineering, such connection needs a common languages and meanings to be able to produce solutions. The paper is aimed at investigating possible solutions to foster the convergence and the dialogue between the two sectors and wants to verify if the complexity of the current situation is changing the concept itself of personalization.
Full-text available
We present our experiences with DatabasePlace, a Web portal aimed at university-level students enrolled in database courses. The portal was established by Addison-Wesley in January 2003. Besides presenting information about the textbooks, the portal also provides additional domain information, online quizzes and three Intelligent Tutoring Systems developed by the Intelligent Computer Tutoring Group (ICTG). We briefly present the three systems, and then discuss our experiences. We also compare the DatabasePlace students to our local students using the three ITSs.
Full-text available
What do we mean by the word ‘context’ in education and how does our context influence the way that we learn? What role can technology play in enhancing learning and what is the future of technology within learning? Re-designing Learning Contexts seeks to re-dress the lack of attention that has traditionally been paid to a learner’s wider context and proposes a model to help educators and technologists develop more productive learning contexts. It defines context as the interactions between the learner and a set of inter-related resource elements that are not tied to a physical or virtual location. Context is something that belongs to an individual and that is created through their interactions in the world. Based on original, empirical research, the book considers the intersection between learning, context and technology, and explores: the meaning of the concept of context and it’s relationship to learning the ways in which different types of technology can scaffold learning in context the Learner-Centric ‘Ecology of Resources’ model of context as a framework for designing technology-rich learning environments the importance of matching available resources to each learner’s particular needs the ways in which the learner’s environment and the technologies available might change over the coming years the potential impact of recent technological developments within computer science and artificial intelligence This interdisciplinary study draws on a range of disciplines, including geography, anthropology, psychology, education and computing, to investigate the dynamics and potential of teacher-learner interaction within a learning continuum, and across a variety of locations. It will be of interest to those teaching, researching and thinking about the use of technology in learning and pedagogy, as well as those involved in developing technology for education and those who use it in their own teaching.
Full-text available
We report the results of a controlled evaluation of an interactive on-line tutoring system for high school math achievement test problem solving. High school students (N = 202) completed a math pre-test and were then assigned by teachers to receive interactive on-line multimedia tutoring or their regular classroom instruction. The on-line tutored students improved on the post-test, but the effect was limited to problems involving skills tutored in the on-line system (within-group control). Control group students showed no improvement. Students' use of interactive multimedia hints predicted pre- to post-test improvement, and benefits of tutoring were greatest for students with weakest initial math skills.
Full-text available
The goal of this paper is to discuss how adaptive search systems which embed exploratory options should be evaluated. We argue that a state-of-the art evaluation of adaptive search systems should follow a "layered evaluation" approach. To support and explain this argument we describe how the layered approach was applied to the evaluation of the adaptive search component of Knowledge Sea II, a system that is powered by a social navigation support mechanism.
In recent years, the learner models of some adaptive learning environments have been opened to the learners they represent. However, as yet there is no standard way of describing and analysing these 'open learner models'. This is, in part, due to the variety of issues that can be important or relevant in any particular learner model. The lack of a framework to discuss open learner models poses several difficulties: there is no systematic way to analyse and describe the open learner models of any one system; there is no systematic way to compare the features of open learner models in different systems; and the designers of each new adaptive learning system must repeatedly tread the same path of studying the many diverse uses and approaches of open learner modelling so that they might determine how to make use of open learner modelling in their system. We believe this is a serious barrier to the effective use of open learner models. This paper presents such a framework, and gives examples of its use to describe and compare adaptive educational systems.
Tutoring systems are described as having two loops. The outer loop executes once for each task, where a task usually consists of solving a complex, multi-step problem. The inner loop executes once for each step taken by the student in the solution of a task. The inner loop can give feedback and hints on each step. The inner loop can also assess the student's evolving competence and update a student model, which is used by the outer loop to select a next task that is appropriate for the student. For those who know little about tutoring systems, this description is meant as a demystifying introduction. For tutoring system experts, this description illustrates that although tutoring systems differ widely in their task domains, user interfaces, software structures, knowledge bases, etc., their behaviors are in fact quite similar.