aied 2015 poster- off the beaten path: the impact of adaptive content sequencing on student...

1
Goal: Guide students to right content in intelligent educa1onal systems Idea: Social guidance based on open social student modeling Allows students to explore each others model or cumula1ve model of the class Increases student engagement Provides effec1ve naviga1on support Mo+va+on Challenges in Social Guidance Adap+ve Sequencing Impact of GS on Learning Gain, Learning Speed, System/Class Performance Greedy sequencing (GS): Aims at maximizing student level of knowledge in domain concepts Off the Beaten Path: The Impact of Adap+ve Content Sequencing on Student Naviga+on in an Open Social Student Modeling Interface Classroom Study R. Hosseini, I.H. Hsiao, J. Guerra, P. Brusilovsky Naviga+onal PaIerns Problem: How to avoid students becoming more conserva1ve with their work with content? Solu+on: Increasing the personaliza1on power of social guidance Combine social guidance with adap1ve sequencing of contents Open Social Student Modeling Students in the class (you are 4th out of 7) 4. Me > User Modeling database Greedy Sequencing Knowledge Report Service Rank C 1 Prerequisites Outcomes Content C 1 : Concepts P: ra1o of known prerequisites O: ra1o of unknown outcomes n p : number of prerequisites n o : number of outcomes Greedy Sequencing Rank Rank = n p P + n o O n p + n o GS and Social Guidance Star size is rela1ve to the rank of content A bigger star means content has higher priority 143 undergraduates in ASU (Fall 2014), in Java Programming and Data Structure course 111 problems — 103 examples — 19 topics Part: (1) No Sequencing (Aug. 21 – Sep. 25) (2) Introduced Sequencing (Sep. 26 – Oct. 21) Logs: 86 subjects — 53 of them had at least 30 problem aaempts Rela1ve Frequencies of topicbased paaerns GS promotes nonsequen1al paaerns 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Weak students Strong students Normalized learning gain Nonfollowers Followers 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 Weak students Strong students % Learning speed Nonfollowers Followers No significant differences in the learning gain Followers with high prior knowledge learn faster (p=.039) Correctness is more frequent in recommended problems (p<.001) Aaemp1ng a recommended content is associated with 0.56 increase in final grade (SE=0.24, p=.017) ~ 9 1mes greater than a not recommended content WithinTopic NextTopic JumpForward JumpBackward Part 1 Part 2N Part 2R 0.08 0.08 0.16 0.68 0.06 0.05 0.12 0.78 0.17 0.17 0.2 0.47

Upload: roya-hosseini

Post on 15-Aug-2015

39 views

Category:

Presentations & Public Speaking


3 download

TRANSCRIPT

Goal:  Guide  students  to  right  content  in  intelligent  educa1onal  systems  Idea:  Social  guidance  based  on  open  social  student  modeling  § Allows  students  to  explore  each  others  model  or  cumula1ve  model  of  the  class  

§ Increases  student  engagement  § Provides  effec1ve  naviga1on  support  

Mo+va+on   Challenges  in  Social  Guidance  

Adap+ve  Sequencing  

Impact  of  GS  on  Learning  Gain,  Learning  Speed,  System/Class  Performance      

Greedy  sequencing  (GS):  Aims  at  maximizing  student  level  of  knowledge  in  domain  concepts  

Off  the  Beaten  Path:  The  Impact  of  Adap+ve  Content  Sequencing    on  Student  Naviga+on  in  an  Open  Social  Student  Modeling  Interface    

Classroom  Study  

R.  Hosseini,  I.H.  Hsiao,  J.  Guerra,  P.  Brusilovsky  

Naviga+onal  PaIerns  

Problem:  How  to  avoid  students  becoming  more  conserva1ve  with  their  work  with  content?  Solu+on:  Increasing  the  personaliza1on  power  of  social  guidance  § Combine  social  guidance  with  adap1ve  sequencing  of  contents  

Open  Social  Student  Modeling  

Students  in  the  class  (you  are  4th  out  of  7)  

4.  Me  -­‐>  

User  Modeling  database  

Greedy  Sequencing  

Knowledge  Report  Service  

Rank  C1  

Prerequisites  Outcomes  

Content  C1:  Concepts  

P:  ra1o  of  known  prerequisites  O:  ra1o  of  unknown  outcomes  np:  number  of  prerequisites  no:  number  of  outcomes  

Greedy  Sequencing  Rank  

Rank =npP + noOnp + no

GS  and  Social  Guidance  

§ Star  size  is  rela1ve  to  the  rank  of  content  § A  bigger  star  means  content  has  higher  priority  

§ 143  undergraduates  in  ASU  (Fall  2014),  in  Java  Programming  and  Data  Structure  course    

§ 111  problems  —  103  examples  —  19  topics  Part:  (1)  No  Sequencing  (Aug.  21  –  Sep.  25)                      (2)  Introduced  Sequencing  (Sep.  26  –  Oct.  21)  Logs:  86  subjects  —  53  of  them  had  at  least  30                          problem  aaempts  

Rela1ve  Frequencies  of  topic-­‐based  paaerns  

§ GS  promotes  non-­‐sequen1al  paaerns    

0  

0.1  

0.2  

0.3  

0.4  

0.5  

0.6  

0.7  

0.8  

0.9  

Weak  students   Strong  students  

Normalize

d  learning  gain  

Non-­‐followers   Followers  

0.00  

0.20  

0.40  

0.60  

0.80  

1.00  

1.20  

1.40  

1.60  

1.80  

2.00  

Weak  students   Strong  students  

%  Learning  speed    

Non-­‐followers   Followers  

§ No  significant  differences  in  the  learning  gain  § Followers  with  high  prior  knowledge  learn  faster  (p=.039)  

§ Correctness  is  more  frequent  in  recommended  problems  (p<.001)  

§ Aaemp1ng  a  recommended  content  is  associated  with  0.56  increase  in  final  grade  (SE=0.24,  p=.017)    ~  9  1mes  greater  than  a  not  recommended  content  

Within-­‐Topic  

Next-­‐Topic  

Jump-­‐Forward  

Jump-­‐Backward  

Part  1   Part  2-­‐N   Part  2-­‐R  

0.08

0.08

0.16

0.68

0.06

0.05

0.12

0.78

0.17

0.17

0.2

0.47

Jump−Backward

Jump−Forward

Next−Topic

Within−Topic

Part 1 Part 2−N Part 2−R