Mash-Up Personal Learning Environments - E-Infrastructures and Technologies for Lifelong Learning - page 133

changed parameters in the environment or context
(Dolog, 2008).
Three important streams of research can be
identified as relevant for personalized adaptive
learning: (1) technologies from adaptive (educa-
tional) hypermedia , (2) learning design technolo-
gies , and (3) adaptive hypermedia generators .
On a finer level, adaptive and intelligent
technologies can be distinguished into cur-
riculum sequencing and problem-solving sup-
port technologies (Brusilovsky, 1999). Whereas
sequencing deals with adapting the navigational
path through pre-existing learning material,
problem-solving support technologies deal with
evaluating student-created content representa-
tions either summatively or - in interactive sup-
port technologies - formatively, even during the
learning process itself, through the provision of
feedback or by presentation of related examples.
Furthermore, in the more generic adaptive hy-
permedia area, adaptive navigation support and
adaptive presentation support can be distinguished
(Brusilovsky, 1999). Both deal with adapting pre-
existing content: adaptive navigation deals with
path and link adaptation (though in a more open
setting - the web), while adaptive presentation
is concerned with the presentation of a subset of
the content in new arrangements to accommo-
date user's needs. Additionally, a third class of
approaches is mentioned by Brusilovsky (1999),
which deals with student model matching: they
try to make use of collaborative filtering aspects
(either by identifying matching peers or by iden-
tifying differences to avoid problems).
Brusilovsky & Henze (2007) identify the
lack of reusability and interoperability as a ma-
jor problem in personalized adaptive learning.
When applying adaptation in the web, this re-
sults in the 'open corpus problem' which can (at
least partially) be compensated by gaining more
interoperability. For adaptation interoperability,
however, standards are still missing (Brusilovsky
& Henze, 2007; Kravcik, 2008).

Holden & Kay (1999) postulate that scrutability
has to become a key characteristic in personaliza-
tion strategies: evidence accreted (i.e. collected)
from various sources is resolved (i.e. assessed)
at request time, while providing control over the
input as well as output streams and inspection ca-
pabilities for the processing mechanisms. Though
this offers triggers for reflective activities, these
are not part of the modelled user activities. They
merely are performed outside the system, thereby
neither supported nor hindered by the system.
Although these adaptive (educational) hy-
permedia technologies all differ, they share one
characteristic: they deal primarily with the naviga-
tion through content, i.e. the represented domain
specific knowledge. Information processing and
knowledge construction activities are not in the
focus of these approaches. Consequently, they do
not treat environments as learning outcomes and
they cannot support learning environment design.
Koper & Tattersell (2005) state that in their
learning design (LD) introduction they will be
using 'learning design' synonymously with 'in-
structional design', though there may be a slightly
different accent in the meaning of both. Specht &
Burgos (2007) elaborate on the adaptation pos-
sibilities in general and particularly within IMS-
LD. However, among the generic components of
educational systems that can be adapted, they list
only pacing, content, sequencing, and navigational
aspects. Neither does the environment (not even
in the sense of tools) appear in this list, nor is it
a driving factor for information gathering, nor
method of adaptation (Specht & Burgos, 2007).
Towle & Halm (2005) discuss how adaptive strate-
gies can be embedded with units of learning by
filtering or reordering resources, changing meth-
ods, slotting learners into roles (and scaffolding
role transitions), or by changing activities. Van
Rosmalen & Boticario (2005) investigate how -
besides design time - also run-time adaptation can
be realized with LD, thereby interfacing LD with
distributed multi-agent systems. They tweak LD
to incorporate agents (by adding them as 'staff',