11. 2: UM adquisition Asking the user (or teacher or …) Observing user behavior Student User Model
12.
13.
14.
15. Inducing from mouse movements Rosa M. Carro, Department of Computer Science, Universidad Autónoma de Madrid Mouse movements Learning styles (ILS) Offline processing sequential global seq global m aximum vertical speed (pixels/ms) m aximum vertical speed (pixels/ms) r = -0.8 accuracy = 94.4%
16.
17.
18.
19. L.S. from Facebook behavior: results Learning style test (n=378) Tree (friends) Tree Fisher (friends) Fisher Active/Reflexive 46,44% 57,00% 60,98% 57,97% Visual/Verbal 52,00% 52,66% 52,41% 37,68% Sensing/Intuitive 40,80% 43,48% 43,09% 41,55% Sequential/Global 60,80% 57,91% 66,67% 62,80%
20. L.S. from Facebook behavior: results If users has posted more than 10 links, more than 6 friends has posted in ther wall during the last year, she has more than 85 friends, is member of at most 13 groups and has more than 34 posts in her wall, then she has a preference for the verbal style
21.
22.
23. Mouse movements in S.N. (iii) Dimension Mouse (n=20) Touchpad(n=13) Active/Reflexive 60,00% 46,15% Visual/Verbal 50,00% 61,54% Sensing/Intuitive 55,00% 53,85% Sequential/Global 35,00% 46,15% Dimension Tool Activity Relevant variables Active/Reflexive Mouse Test, text Mean horizontal acceleration Max vertical acceleration Visual/Verbal Touchpad Sorting Time used Sensing/Intuitive Mouse T/F T/F Min vertical acceleration Max vertical acceleration Sensing/Intuitive Touchpad Browsing Max vertical acceleration
39. Adapting to LS: an example ILS VALUE ON SEQUENTIAL/GLOBAL: Extreme and mild Sequential Well balanced Extreme and moderate Global
40. Adapting to LS: an example ILS VALUE ON SEQUENTIAL/GLOBAL: Extreme and mild Sequential Well balanced Extreme and moderate Global
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53. Visualizing students Visualization of the 2 dimensions that showed to be the most relevant in previous research values for dimension 1 values for dimension 2
63. Thanks! Questions? 7th International Workshop on Authoring of Adaptive and Adaptable Hypermedia [email_address]
Notas del editor
Concept learning is the induction of a concept (category) from observations Deductive arguments are attempts to show that a conclusion necessarily follows from a set of premises or hypotheses
18 estudiantes We found a strong correlation between maximum vertical speed and sequential/global dimension score. In addition, it was possible to predict this dimension of the students’ learning style with high accuracy (94.4%, correlation coefficient r = - 0.8). This suggests that mouse movement patterns can be a powerful source of information for user model acquisition.
Se capturan las coordenadas del cursor cada 40 secs
So, if the teacher (course designer) wants to implement some form of cooperative learning, on key issue is, of course, divide the people in groups.
I may be the case where groups built following some rational criteria would produce better results. E-learning: usually people don ’t know one to the other, and coordination use to be more difficult than in a face2face environment
If we were able to built groups following some “intelligent” criterion, which would be such a criterion? With “heterogeneous” groups I mean groups with diversity, with people with different ...
On the most fascinating books I have read ever ... In this books you can read statements like: Even if one of the best, it ’s not the only work to appraise the benefits of diversity
So the goal is diversity. I mean: building groups with people as diverse as possible. What you take into account (from the student profile) to measure diversity can change. Even if our focus is on L.S. the techniques can apply to any measurable dimension of the student personality
El problema es explicarlo sin fórmulas ni gráficos
We can think on genetic algorithms
is behind the name of the fifth track from U2's 1993 album, Zooropa,
Good results are more likely to be obtained if we could try different configurations
So, with the goal of trying different setups, is when TOGETHER comes to scene Basically, the tool executes the algorithm 100 times with different order of students, and store the best solution. Now, WAIT A MOMENT! What about “it’s no possible to decide which is the best solution” story I was telling?
Now, how does TOGETHER visualize this information? Remember that visualization is an aid for the final decision, and because of that showing only a two most significant dimensions seems a sensible decision. Each dot represents a student, and the position of the dot represents the respetive L.S. dimension values.
Hablar del significado de los colores
Then we have two numbers (describing somehow the group), that we can use in a new, different visualization. The numbers on each dimension are sum of distances, and of course they are always positive numbers
When we put together the information about all the groups, we have this chart. Colors are used to show how much groups share the same point in the “distances’ space”
In this chart, Up and right is good, down and left is bad