CYBERLEARNERS AND LEARNING RESOURCES

Leyla Zhuhadar, PhD

- Adj. Assistant Prof. CECS Dept., University of Louisville, USA.- Research Scientist, WKU, USA.

Rong Yang, PhD (Speaker)

- Assistant Professor, Department of Mathematics and Computer Science, Western Kentucky University, USA.

• How can we detect a community of similar Cyberlearners based on the structure of a huge social network?

• How can we present this interconnection among communities visually to analyze our Cyberlearners’ behaviors?

• Finally, building a community-based recommendation system.

THE MAIN THEMES

Social Learning Network Analysis (Buckingham & Ferguson, 2011)

CYBERLEARNER AND OPEN SOURCE PLATFORMS

1. Technology enhanced learning: Open-source educational resources (any place, any time, and in any way)

2. Using state of the art data mining algorithms and Web services

3. Adopting a learner-centered pedagogical approach

4. Offering a mix of diverse content via Web 3.0.5. Providing metadata, semantic, visualized, and

cross-language searchable content. 6. Recommender System

HYPERMANYMEDIA

IS ALIGNED WITH THE FOLLOWING IDEAS:

HyperManyMedia Repository: http://hmm.wku.edu

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

WHAT IS THE HYPERMANYMEDIA REPOSITORY?

CYBERLEARNERS’ PREFERENCES!

1% 3%7%

12%

5%

6%

8%

58%

Cyberlearners' Preferences (2006-2011)

Generic Search Metadata SearchSemantic Search Personalized Semantic SearchPersonalized Search with User Relevance Feedback Collaborative FilteringCross-Language Search Visual Search

> 750,000 Cyberlearners

• Hierarchical Arrangement of:

HYPERMANYMEDIA RESOURCES

Colleges (14)

Courses(60)

Resource(838)

Audio, Video, Lecture, RSS feed, Podcast, Vodcast, etc.

( >10,000)

• This is great! But is our cognitive system able to deal with this vast amount of resources? • The most difficult question raised here: “Is our conceptual recognition of these learning resources able to find what we really want?

• In 1956, George Miller discovered the magic number:• 7 +/-2 = limited capacity of

our Short Term Memory• Digital span, letter span, and

visual matrix

THE MAGIC NUMBER OF SHORT TERM MEMORY (STM)

REMINDER:

I am a Cyberlearner and need help to find a learning resource!

But, I really don’t know what type of help I need!

LINKING SOCIAL NETWORKS WITH RECOMMENDER SYSTEM: WHO ARE MY NEIGHBORS?

I am a Cyberlearner and need help to find a resource!

But I really don’t know what type of help I need!

• Yes! We provided our Cyberlearners with a semantic recommender system that gives them related resources to their search; but is this enough?

• Can I help our Cyberlearners to remember these learning resources by linking/relating them conceptually to other resources?

THE MAGIC NUMBER OF STM (7+/-2)

• But, how can we find this community with common• Learning domains,• Problems,• Interests, and• Learning styles?

• Especially, when we have a system of thousands of resources and hundreds of thousands Cyberlearners navigating. We really need help!

SEARCHING FOR ANSWERS?

• Proposing a bottom-up approach (No pre-knowledge).

• Data-driven approach: archived activities of Weblogs for the last 6 years of Cyberlearners visited HMM (~750,000).

• Looking underneath the structure of HMM social networks.

SEARCHING FOR ANSWERS?

This graph represents a social network structure of a weblog (2/1/2011- 8/1/2011). ~8,000 Cyberlearners and ~24,000 (edges) connections among learners and resources in HMM.

• Network with High Complexity• Small world (Kleinberg, 2000)• Mine the structure to of the network to

answer the posed question• Reminder! Simplistic approach • Modularity measurement was used to

visualize the network structure.

; where represents the weight of the edge between i and j, = is the sum of the weights of the edges attached to vertex i, is the community to which vertex i is assigned, the δ-function δ(u, v) is 1 if u = v and 0 otherwise and m =.

FINDING COMMUNITY!

• Discovering the community of Cyberlearners; Each dot in this graph is a learner.

• 10 communities of learners with similarity (commonality).

• Of course the distribution among the number of dots ( Cyberlearners) varies; for the sake of simplicity, we assume they are equally distributed.

FINDING COMMUNITY!

• If I am a Cyberlearner, I definitely belong to one of these communities. Therefore, instead of being a dot among 8,000 dots, I am now a dot among 800 dots: Still it is a huge number

• If I need a recommendation, I don’t want to receive help from 800 Cyberlearners in my community!

FINDING COMMUNITY!

• Observing the graph (carefully): • Each Cyberlearner has a unique

distance from the hub. • A dot ahead is another learner (a

little bit more experienced with the resources in this domain - closer the hub).

• A dot behind is a learner less experienced.

• A learner very close to the hub could be considered an expert.

FINDING COMMUNITY!

1. Do we want to intimidate a Cyberlearner with an expert?

2. Or, do we provide the Cyberlearner with the learner closest to him/her?

• distance-based = who has the most similar profile to him/her

FINDING COMMUNITY!

Our answer is neither one!• We used another concept in

cognitive psychology—Chunking Hypothesis.

• In 1978, Herbert Simon introduced the chunking hypothesis and won a Nobel Prize in economics. "for his pioneering research into the decision-making process within economic organizations" (1978).

FINDING COMMUNITY!

• Holding the concept of a primitive set (Magic Number) and the concept of chunking;

• Magic Number: Each Cyberlearner is recommended with resources he/she did not visit before from his/her closest 3 neighbors (triangle); and

• Chunking: those recommendations should range from 5 to 9 (no more).

CONCLUSIONS

This graph represents a social network structure of a weblog (May-October 2011). ~8,000 Cyberlearners and ~24,000 (edges) connections among learners and resources in HMM.

This graph represents a social network structure of a weblog (May-October 2011). ~8,000 Cyberlearners and ~24,000 (edges) connections among learners and resources in HMM.

This graph represents a social network structure of a weblog (May-October 2011). ~8,000 Cyberlearners and ~24,000 (edges) connections among learners and resources in HMM.

DID WE CONNECT THE DOTS?

I am a Cyberlearner and need help to find a resource!

But I really don’t know what type of help I need!

I am a Cyberlearner and need help to find a resource!

But I really don’t know what type of help I need!

DID WE CONNECT THE DOTS?

HOW?

Open, social learning Open Universities (Open Universities (UK, Germany,

India, etc.) Open Courseware (MIT, Khan Academy, etc.) Large open online courses ( Stanford: AI & ML)

Social Learning Analytics Social learning network analysis Social learning discourse analysis Social learning content analysis Social learning disposition analysis Social learning context analysis

THE FUTURE OF CYBERLEARNERS

1. Simon Buckingham Sum and Rebecca Ferguson, Social Learning Analytics, Knowledge Media Institute, Social Learning Analytics, 2011.

2. George Miller (Magic Number, 1956)

3. Phil Long and George Siemens, Penetrating the Fog: Analytics in Learning and Education, 2008.

4. Small-World Phenomena and Decentralized Search: Kleinberg. Navigation in a Small World. Nature 406 (2000), 845.

5. Herbert Simon, The chunking hypothesis, http://en.wikipedia.org/wiki/Herbert_Simon, 2005.

6. Mattieu Latapy, Main-memory triangle computations for very large (sparse (power-law)) graphs, 2010.

REFERENCES

THANKS FOR YOUR ATTENTION!

Rong Yang, PhD. (speaker)Email: rong.yang@wku.edu

&Leyla Zhuhadar, Ph.D.

Email: leyla.zhuhadar@wku.edu