Towards Generic Adaptive Systems: Analysis of a Case Study

1
Towards Generic Adaptive SystemsAnalysis of a
Case Study
Databases Hypermedia Group, Department of
Informatics

• Licia Calvi Alexandra Cristea

AH02 May 29-31, 2002, Malaga
2
Keywords

• Generic AS
• XML
• AHA!

3
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

4
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

5
Ideas Motivation

• AH author can separate w. difficulty
• links vs. concepts
• adaptive navigation vs. a. presentation
• carefully design a synchronous system
• a better way to look at AH authoring pb
• combination of CM paradigm for course narrative
• several new adaptation rules

6
Method

• analyze AHA!
• for general observations on AHS
• improvement suggestions for AHA!
• suggest a concept-based AHS structure
• extend rule-based overlay method for
  user-adaptation
• another step towards flexible generic-purpose AH

7
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

8
A few words on AHA!

• well-known system
• AH pioneer (1st version 1996/97)
• domain benchmark
• power popularity due to simplicity

9
AHA! adaptation methods

• ? page concept, showed/not acc. to conds (on
  vars) in XML file (requirement list)
• vars changing rules ? simple (generate list)
• cond. fragm. in pgs AHA tag language (XML based)
• 1-2 adaptive navigation support (pg level)
• 3 adaptive presentation

10
Problems

• Lack of
• Reusability
• Expressivity

11
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

12
Concept mapping

• intuitive classif. divide source material into
  concepts ? piece has independent semantics (
  semantic Web)
• low level atomic concepts
• concepts collections composite concepts
• together concept hierarchy
• primitive building blocks of hypermedia

13
Linking

• building blocks w. diff. sequences ? diff.
  presentations (high granularity level concept
  level)
• adaptive navigation support
• adaptive presentation at lower, concepts
  fractions level
• E.g. text intro. can be used w. other
  introductory fragments in introductory chapter,
  (to drop at later browsing) etc.
• Pb. no independent meaning.
• common solution
• divide concepts into sub-concepts but
• pb semantics loss collaborative authoring
  cannot be semantically annotated not
  significant for search mechanisms.

14
Attributes

• more appropriate
• concept name, alternative contents, fragments,
  etc.
• course content mapped on concept hierarchy
  describing concepts w. attributes set, adaptation
  concept-level attribute adaptation.
• Advantage can be performed viewed from high
  level
• no need of separate consideration of cond.
  fragments in texts (difficult to re-use by other
  authors)
• content - adaptative engine rules authoring is
  separated ? easier automatic checks
• adaptation combining concept attributes into
  pages
• (info pieces that can be show at a time)

15
Navigation

• dependent on presentation format
• e.g. a handheld device w. short pg displays
  next button more often within same lesson
• model compatible w. RDF standard
• resources ? concepts,
• properties ? attributes
• literals ?attribute values

16
Adaptive navigation presentation
17
What is already in AHA?

• main difference
• Concepts at pg granulation
• pgs constructs (conditional fragments) ? concept
  attrs cannot be independently used w. other
  concepts or c. attributes.
• under development version consider multiple
  attributes a DB structure, that allows
  flexibility

18
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

19
Typical adaptivity

• Most AS rule-based, i.e.
• Adaptation conditional rules
• IF ltPREREQUISITEgt THEN ltACTIONgt

20
New adaptation rules proposed
21
A level rule

• IF ENOUGH(ltPREREQUISITESgt) THENltACTIONgt
• ENOUGH fct. of no. quality of prerequisites
  true if, e.g., a given no. of prerequisites from
  a set is fulfilled
• Ex PREREQUISITES time_spent ACTION go to
  next level
• Rule becomes
• IF ENOUGH (time_spent on crt. level) THEN go to
  next level
• Where ENOUGH is defined, e.g., as follows
• ENOUGH (time) 30 time units
• time (advanced topic) 10 (time units per
  topic)
• ENOUGH (medium topic) 5 (time units per
  topic)
• ENOUGH (beginner topic) 2 (time units per
  topic)

22
A temporal rule

• action repeated as long as 1-more cond.s hold
• WHILE ltCONDITIONgt DO ltACTIONgt
• E.g warning - user search direction is wrong ?
  service denial over a threshold / drill ex.

23
A repetition rule

• a certain (simple / composed) action repeated for
  a no. of times predefined by author
• FOR lti1..ngt DO ltACTIONgt
• E.g. time action has to last before reader can
  move on.

24
An interruption command

• forced user to do smthg. else
• BREAK ltACTIONgt
• exacerbation of traditional AHS behavior user
  punished for not sticking to learning pathways

25
A generalization command

• new concept reached is compared w. more general
  ones it refers to. As a result, the reader is
  pointed to related concept(s)
• GENERALIZE (COND, COND1, , CONDn)

26
A specialization command

• if concept is general, system deductively points
  reader to more specific instantiations
• SPECIALIZE (COND, COND1, , CONDn)
• E.g, if student reads about Model Reader in a
  course on postmodern literature, she can be
  pointed to an extract from Calvinos novel Se
  una notte, where this notion is exemplified.

27
Other commands

• comparison (concept analogy search)
• difference
• both instances of generalization
• duration a rule related to repetition
• lyrical use of repetitions in hyperfiction has
  given rise to a particular design pattern

28
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

29
(No Transcript)
30
(No Transcript)
31
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

32
Pros cons

• Balance system complexity vs. authoring
  efficiency
• Extending AS w. extra adaptation rules is
  beneficial if rules can express situations that
  were not possible (difficult) to express w. given
  set of tools/ rules.
• Makes sense if it doesnt increase dramatically
  the types of tests an AH author has to do to
  verify his/her output.

33
Future directions

• automatic check tools
• visual checking mechanisms, dynamical
  representation of processes involved
• E.g., effect of new rule on rest can be shown on
  static (smaller) link graph, as a propagation of
  a colored fluid through graph, etc.

34
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

35
Conclusion

• suggested a better approach for AH authoring
• combination of CM paradigm to construct course
  narrative several new adaptation rules.
• highlighted new rules to integrate into AH
  authoring shell / toolkit.
• showed integration of 2 formalisms in AHA! ex.
  version (for more adaptivity)
• We claim that this approach is another step
  towards flexible generic-purpose adaptive
  hypermedia.

36

• Thank you for listening!

37
Recent AHA! extensions

• editor to connect requirements to pages
• editor for generate rules
• forms to make changes to UM
• most important form allowing AH user to modify
  knowledge attributes assoc. to page-concepts.

38
System complexity vs. authoring efficiency
39
Index

• Motivation background
• AHA! beyond
• Concept-mapping paradigm
• New Adaptation Rules How to Augment the
  Adaptation Engine
• Implementing New Rules in the current AHA!
• Problems need of checking mechanisms
• Future directions
• Conclusion

40
Problems need of check mechanisms

• by increasing system complexity, authoring
  efficiency grows for a while, and then drops.
• AHA! is somewhere at beginning of slope
• Adding more features flexibilities can increase
  authoring efficiency for a while, but how to stop
  before down-curve?
• E.g., when authors deal w. complex graphs w. many
  concepts attr.s, its easy to leave something
  out by mistake.

41
AHAM

• tries to deal w. such pb. as
• termination ( avoiding of loops)
• confluence (equivalence of rule execution order)
• T activation graphs (active DBs static
  analysis)
• possible states graph det. by concepts, links,
  attr.s, values (init. val.s ranges search
  tree constrains optimization) rule sets
• If graph has no loops, system will always
  terminate.
• C commutation check (rule pairs order
  equivalence)
• AHA! only monotonic attributes (per concept)
  increase ? termination but difficult in next
  version w. multiple attr.s
• AHA! doesnt deal w. confluence.

42
Other problems

• concepts (or concept fragments) never reached
• rules (or other adaptation mechanisms) w.
  attributes w. out of range (or domain) values.

43
New rules

• good news dont require extra checking
  mechanisms
• ? loops in regular rules will also ? in level -,
  temporal -, repetition rules.
• Non-equivalent non-commutable rules to be
  executed at a given time pose same problems on
  extended set.
• Extended commands of generalization
  specialization can be treated the same as regular
  links.
• Interruption command can help in breaking
  infinite loops, Java catch-throw mechanism of
  exception handling.

44
New rules

• bad news time - space -consuming.
• better way simplifications complexity
  decreasing assumptions.
• E.g., belief revision technique to check
  inconsistencies in knowledge attr.s to concepts
  consequent knowledge acquisition pb.

45
Belief revision

• introduction of a case-based heuristics that
• recalls previous concept w. same features
  assoc. attributes
• adapts course struct., via rule-based formalism,
  to current learning scenario
• resolves inconsistencies so that changes of state
  are epistemologically conservative (resulting
  narrative is not subverted).

46
Future directions

• W. standardiz. of AS building bricks (LOM,
  Learner model IEEE, LTSC for education, RDF,
  etc.) its feasible to collaborate share
  adaptive techniques, technologies also system
  parts, AH presentations, etc.
• Adaptive adaptable systems
• are necessary in education, where learners come
  w. different cultural knowledge backgrounds,
  learning styles, genders, ages, (context
  life-long learning).
• are definitely necessary in commerce
  (Amazon.com).
• can have surprising applications adaptive
  literature art.

47
Conclusion distructive

• criticized widespread practice to distinguish
  adaptation in hypermedia between adaptive
  navigation support adaptive presentation
  because
• AH authors have to artificially separate links
  from concepts but still coordinate them to
  provide a conceptually valid adaptation that
  contributes to a significant knowledge
  acquisition.