1. Mash-Up Personal Learning Environments TENcompetence Winter School, February 2nd, 2009, Innsbruck Fridolin WildVienna University of Economics and Business Administration
3. Structure of this Talk Preliminaries Critique: Flaws of Personalisation Personal Learning Environments (F.I) End-User Development (F.II) Activity Theory (F.III) A Mash-UpPLE (MUPPLE.org) The Rendering Engine The Scripting Language The Prototype An Example Activity Sharing Patterns Conclusion Problem Fundamentals Solution
5. ... Are probably around us ever since the ‚homo habilis‘ started to use more sophisticated stone tools at the beginning of the Pleistocene some two million years ago. Learning Environments = Toolsthat bring together people and content artefacts in activities that support in constructing and processing information and knowledge.
6. Assumption I. Learning environments are inherently networks: encompass actors, artefacts, and tools in various locations with heterogeneous affiliations, purposes, styles, objectives, etc. Network effects make the network exponentially more valuable with its growing size
7. Assumption II. Learning Environments are learning outcomes! not an instructional control condition! For example, a learner may prefer to email an expert instead of reading a paper: Adaptation strategies go beyond navigational adaptation through content artefacts Setting up and maintaining a learning environment is part of the learning work: future experiences will be made through and with it knowing tools, people, artefacts, and activities (=LE) enables
8. Assumption III. Learning to learn, while at the same time learning content is better than just (re-) constructing knowledge. Acquisition of rich professional competences such as social, self, and methodological competence ... is superior to only acquiring content competence (i.e. Domain-specific skill, facts, rules, ...) Due to ever decreasing half-life of domain-specific knowledge Construction != Transfer!
9. Assumption IV. Designing for Emergence ... is more powerful than programming by instruction Emergent behaviour: observable dynamics show unanticipated activity Surprising: the participating systems have not been instructed to do so specifically (may even not have intended it) Why? Because models involved are simpler while achieving the same effect Example: Walking Robot
11. Flaws of Personalisation Claim: Instructional design theories and adaptive & intelligent technologies do not support or even violate these assumptions!
13. Instructional Design Theories ... offer explicit guidance to help people learn better But: Environment = instructional control condition (cf. e.g. Reigeluth, 1999) But: Environment = separate from desired learning outcomes (cf. e.g. Reigeluth, 1999) Even in constructivist instructional theories: LE is created by instructional designer (cf. e.g. Mayer, 1999; Jonassen, 1999) Appear in applied research in two flavours: with and without strong AI component
14. Strong AI Position = system intelligence monitors, diagnoses, and guides automatically Inherently ill-defined:cannot monitor everything Constantly overwhelmed:what is relevant Computationally expensive:or even impossible Even if: no understanding (cf. Searle’s Chinese Room) (from: modernlove.comicgenesis.com)
15. Weak AI Position mixture of minor automatic system adaptations along a coarse-grain instructional design master plan engineered by a teacher or instructional designer Learning-paths are fine-tuned along learner characteristics and user profiles to conform to trails envisioned (not necessarily proven) by teachers But: No perfect instructional designer In fact: most instructors are only domain-experts, not didactical ones
16. Weak AI Position (2) Furthermore: planned adaptation takes away experiences from the learner: External planning reduces challenges Thus reduces chances to become competent Learners are not only sense-makers instructed by teachers along a predefined path Learners need to actively adapt their learning environments so that they can construct the rich professional competences necessary for successful learning (cf. Rychen & Salganik, 2003)
17. Instructional Design Theories Locus of control only with the instructional designer or with the system Not (not even additionally) with the learner But: Learners are not patients that need an aptitude treatment. => Shortcoming of ID Theory!
19. Adaptation Technologies Varying degree of control: Adaptive ←‒‒‒‒ fluent segue ‒‒‒‒-> AdaptableSystem adapts ←‒‒‒‒‒‒‒‒‒‒-> User adapts(Oppermann, Rashev, & Kinshuk, 1997; Dolog, 2008) Three important streams: Adaptive (educational) hypermedia Learning Design Adaptive Hypermedia Generators
20. Adaptive (Educational) Hypermedia Generic Types: adaptive navigation support: path and link adaptation adaptive presentation support: presentation of a content subset in new arrangements Education Specific Types: Sequencing: adaptation of the navigation path through pre-existing learning material Problem-solving: evaluate the student created content summatively or formatively through the provision of feedback, etc. Student Model Matching:collaborative filtering to identfy matching peers or identify differences (Brusilovsky, 1999)
21. Adaptive (Educational) Hypermedia Main Problems of AEH: Lack of reusability and interoperability Missing standards for adaptation interoperability primarily navigation through content (=represented domain-specific knowledge) Processing and construction activities not in focus Environments are not outcomes, do not support environment design (cf. Henze & Brusilovsky, 2007; Holden & Kay, 1999; Kravcik, 2008; Wild, 2009)
22. Learning Design Koper & Tattersell (2005): learning design = instructional design Specht & Burgos (2007): adaptation possibilities within IMS-LD: Only pacing, content, sequencing, and navigational aspects environment is no generic component that can be adapted (or tools/functions/services), nor driving factor for informaiton gathering nor method for adaptation Towle and Halm (2005): embedding adaptive strategies in units of learning
23. Learning Design Services postulatedtobeknownat design time (LD 1.0 hasfourservices!) Services havetobeinstantiatedthrough formal automatedprocedures But: Van Rosmalen & Boticario: runtime adaptation (distributed multi-agents added as staff in the aLFanet project) But: Olivier & Tattersall (2005): integratinglearningservices in theenvironmentsectionof LD
24. LD continued Targets mainlyinstructionaldesigners(seeguidelines, seepractice) But: Olivier & Tattersall (2005) predictapplicationprofilesthatenhance LD withserviceprovidedbyparticularcommunities, thoughinteroperabilitywithotherplayersthanisnolongergiven But: Extensionsproposed (cf. Vogten et al., 2008): formalisation, reproducability, andreusabilityof LDs can also becatalyzedthroughthe PCM thatfacilitatesdevelopmentoflearning material throughthelearnersthemselves.
25. LD Shortcomings Services != Tools Perceivablesurfaceof a toolmakes a difference (cf. e.g. Pituchand Lee (2004): theuserinterfaceoftoolsinfluencestheprocessespursuedwiththem Agreement on sharingservicescanalwaysonlybethesecondstep after innovatingnewservices Specifyingservicesat design time is inflexible
26. Adaptive Hypermedia Generators LAG: language for expressing information on assembly, adaptation and strategies plus procedures of intelligent adaptation applications Hypertext Structure Rule-based path adaptation (Cristea, Smits, & De Bra, 2007)
27. Adaptive Hypermedia Generators WebML + UML-Guide: client-side adaptation of web applications (Ceri et al., 2005) WebML: follows hypertext model UML-Guide (modified state diagrammes): user navigation through a system can be modelled Both together can generate personalised apps But: restricted to content and path design, And: expert designer recommended
28. Summary of the Critique The prevailing paradigm is ‚rule‘, not ‚environment‘! Learners are executing along minor adaptations what instructional designers (mostly teachers) have foreseen. No real support for learning environment design (= constructing and maintaining learning environments).
30. Personal Learning Environments Not yet a theory and no longer a movement In the revival of the recent years: starting as opposition to learning management systems Common ground: all projects envision an empowered learner capable of self-direction for whom tightly- and loosely-coupled tools facilitate the process of defining outcomes, planning their achievement, conducting knowledge construction, and regulating plus assessing(van Harmelen, 2008)
31. History of PLEs Early Work: Focus on user- and conversation centred perspective (Liber, 2000; Kearney et al., 2005) personal space used for developmental planning and aggregating navigational as well as conversational traces Next Phase: interoperability issues (RSS/ATOM, service integration via APIs, …) (Downes, 2005; Wilson, 2005; Wilson, 2005; Wilson et al., 2007) Today: heterogeneous set of implementation strategies
32. PLE Implementation Strategies Coordinateduse e.g. with the help of browser bookmarks to involved web apps Simple connectors for data exchange and service interoperability Abstracted, generalised connectors that form so-called conduits e.g. those supported by the social browser Flock e.g. by the service-oriented PLE Plex (Wilson et al., 2007)
33. Augmented Landscapes: VLE+PLE individualsuse subsets oftools and services providedby institution actors can choosefrom a growingvariety of options gradually transcendinstitutional landscape actors appear asemigrants orimmigrants leave and joininstitutional landscape for particular purposes
35. End-User Development Deals with the idea that end-users design their environments for the intended usage Evolve systems from ‘easy to use’ to ‘easy to develop’ For example: Excel Scripting Forexample: Apple Script
36. End-User Development Shifting the locus of control from developer to (power) user Coming from modern project management and software development methods (agile, XP, ...) Via User-centred design from HCI: dates back at least to the 1970ies: dedicates extensive attention to the user in each step of the design process, but no development ... and a rather recent research stream (cf. Lieberman et al., 2006)
37. Mash-Up? The ‘Frankensteining’ of software artefacts and data Opportunistic Design (Hartmann et al., 2008; Ncube et al., 2008) ‘Excel Scripting for the Web’ Various Strategies (cf. Gamble & Gamble, 2008)
38. End-User Development Let’s activate the long tail of software development: let’s develop applications for five users!
40. ActivityTheory Structuring Change with Activities Activity is shaped by surroundings E.g. tools have affordances (like a door knob lends itself to opening) Activity shapes surroundings! Activities can result in construction of a tool Long tradition (Leont’ev, 1947; Scandinavian AT: Engeström, 1987)
43. Web-Application Mash-Up { do s.th. } { for an output } share bookmarks { using http://… } using distance.ktu.lt/scuttle RSS feed summarize papers using teldev.wu-wien.ac.at/xowiki find papers using www.objectspot.org
44. Mash-Up PLE (MUPPLE) PLE: change in perspective, putting the learner centre stage again, empower learners to construct and maintain their learning environment Mash-Ups: Frankensteining of software artefacts and data
45. Mash-Up PLE (MUPPLE) Set of Web-Based Tools for learning,client-sided aggregation (= ‘web-application mashup’) Recommend tools for specific activities through design templates through data mining Scrutable: give learner full control over learning process Track learner interaction & usage of tools and refine recommendations “Mupples were small furry creatures that were imprisoned at the Umboo Lightstation when Mungo Baobab, C-3PO and R2-D2 rescued them. Some considered Mupples a delicacy.” -- http://starwars.wikia.com/wiki/Mupple Mash-UP Personal Learning Environments
49. Rendering Engine OpenACS module based on XoWiki and Prototype Windows library Combine tool mashup and Wiki content Provide templates for pre-defined learning activities
51. LISL Design Decisions Natural Language Like, Learnabilitylearners do not need to know a lot about the syntax Extensibilitylearners may define and use own actions Semantics, Recommendationsfor each activity the system offers a landscape of tools Scrutability, Controllabilitylearners receive information about system decisions,but can always change and customize Interoperability, Exchangeabilitylearners can export parts of their ‘learning script’ to hand it over to others Loggingtool interactions can be tracked using ‘invisible’ logging commands
64. Conclusion Learning environments and their construction as well as maintenance makes up a crucial part of the learning process and the desired learning outcomes. Learning environment design is the key to solve shortcomings of today’s theory and practice. ... and mash-up personal learning environments are one possible solution for this.