Keynote for V Jornadas Iberoamericanas de Interacción Humano-Computador 2019
http://mexilab.com/V_Jornadas_HCI/#1#schedule
Drawing on examples from academic research in the field of artificial intelligence in education (AIED) and learning analytics (LA), I will cut through the current hype and make a case for carefully designed systems for a wide range of pedagogies.
As an introduction to the field, the talk will first share how the growing concerns about the role of AI in society, big data and big companies are entering education.
Using the case of exploratory learning, I will then offer possible responses challenging designers, developers and educators, to seize the opportunities afforded by the emerging technological context around data, analytics and AI, while carefully considering design choices when it comes to practical implementation.
Russian Call Girls in Andheri Airport Mumbai WhatsApp 9167673311 💞 Full Nigh...
Artificial Intelligence and Data Analytics in Education: the case of exploratory learning
1. Artificial Intelligence and Data Analytics in Education
The case of exploratory learning
Dr Manolis Mavrikis
UCL Knowledge Lab
@mavrikis
@uclknowledgelab#VJornadasIHC
17. Types of AI in Education
Intelligent Tutoring Systems (ITS)
∙ Break problems into steps
∙ Provide scaffolding and feedback during problem-solving
∙ Adapt content and personalise the experience of learners
Examples online
34. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 731724.
http://iread-project.eu iRead Project @iread_project
38. Exploratory learning
• Targeting mostly conceptual learning
• Criticised for lack of efficiency (Mayer, Kirschner et al. etc.)
• Not the only approach c.f. Rummel et al. ICLS 2016
• Adoption issues due to ‘orchestration’ difficulties (Dillenbourg, 2010)
Rummel, N., Mavrikis, M., Wiedmann, M., Loibl, K., Mazziotti, C., Holmes, W., & Hansen, A. (2016). Combining Exploratory
Learning with Structured Practice to Foster Conceptual and Procedural Fractions Knowledge. In ICLS 2016
Dillenbourg, P., & Jermann, P. (2010). Technology for classroom orchestration. In New science of learning (pp. 525-552). Springer
New York.
38
49. Pedagogic strategies for student support
• Supporting processes of exploration
• Supporting students to set and work towards explicit goals.
• Directing students’ attention.
• Helping students organise their working environment.
• Provoking cognitive conflict.
• Encouraging alternative solutions.
• Supporting reflection
• Promoting motivation
• Supporting collaboration
Mavrikis, M et al. (2008) "Revisiting pedagogic strategies for supporting students’ learning in mathematical microworlds."
Proceedings of the International Workshop on Intelligent Support for Exploratory Environments at EC-TEL.
http://ceur-ws.org/Vol-381/paper04.pdf 49
@mavrikis#VJornadasIHC
52. FRAME - Separation of concerns
Gutierrez-Santos S, Mavrikis M, Magoulas G (2012) A separation of concerns for engineering intelligent support for
exploratory learning environments. Journal of Research and Practice in Information Technology 44(3):347–360
52
Microworld/Model & Events
Analysis
Reasoning
Feedback
53. Intelligent Support
Janning, R., Schatten, C., Schmidt-Thieme, L.: Perceived task-difficulty recognition
from log-file infor- mation for the use in adaptive intelligent tutoring systems. Int. J.
Artif. Intell. Educ. 26(3), 855–876 (2016)
Analysis
Reasoning
Feedback
54. Analysis
Grawemeyer, B., Mavrikis, M., Hansen, A., Mazziotti, C., Gutierrez-Santos, S.
(2014) Employing Speech to Contribute to Modelling and Adapting to
Students' Affective States. EC-TEL 2014. 54
Feedback
Reasoning
56. Feedback
56
Reasoning
Feedback
Grawemeyer, B., Mavrikis, M., Holmes, W. et al. Affective learning: improving engagement and enhancing learning
with affect-aware feedback. UMUAI (2017) 27: 119. doi:10.1007/s11257-017-9188-z http://bit.ly/affect-italk2learn
57. Student Needs Analysis
• Tailor the next exercise to a student based on their:
• Previous task and representations
• Performance on current task
• Level of challenge
• Affective state
@mavrikis @uclknowledgelab 57
58. AI as assistance to human intelligence
• Delegate responsibility in support
• Domain-specific and affect-based feedback
• But by no means aimed to replace teachers !
@mavrikis @uclknowledgelab 58
59. Promising results
• Meta-analyses show impact of intelligent tutoring systems (VanLehn,
2011; du Boulay, 2016)
• Combination of exploratory and structure - Rummel et al. (2016)
• Affect-aware support contributes to reducing boredom and off-task
behavior, and may have an effect on learning (UMUAI, 2017)
Grawemeyer, B., Mavrikis, M., Holmes, W. et al. Affective learning: improving engagement and enhancing learning
with affect-aware feedback. UMUAI (2017) 27: 119. doi:10.1007/s11257-017-9188-z http://bit.ly/affect-italk2learn
59
60. Limitations
• Domain- and Task-specific
• Costly – what about scaling up or genaralising?
• Inherent ‘limits’ of AI
@mavrikis @uclknowledgelab 60
62. • Lack of awareness & ‘control’ of the classroom
ELE Orchestration
Mavrikis, M., Gutierrez-Santos, S., Geraniou, E., Noss, R., & Poulovassilis, A. (2013). Iterative Context Engineering to Inform the Design of
Intelligent Exploratory Learning Environments for the Classroom. In R. Luckin, S. Puntambekar, P. Goodyear, B. L. Grabowski, J. Underwood, N. Winters (Eds.), Handbook of Design in Educational
Technology (pp. 80-92). Routledge.
63. • Could we design tools to assist teachers in their role as facilitators in
classrooms with exploratory environments?
Our challenge
64. Learning Analytics as an answer to AI limits
• Design should be based on analysis of teacher needs
(in the context of AIED systems)
• Where are the ‘actionable insights’ in LA?
@mavrikis @uclknowledgelab 64
66. R. Martinez-Maldonado, A. Pardo, N. Mirriahi, K. Yacef, J. Kay, and A. Clayphan. The LATUX workflow: Designing and deploying awareness tools in
technology-enabled learning settings. In Proceedings of the Fifth International Conference on Learning Analytics And Knowledge, pages 1–10, 2015.
Mavrikis, M., Gutierrez-Santos, S., Geraniou, E., Noss, R., & Poulovassilis, A. (2013). Iterative Context Engineering to Inform the Design of
Intelligent Exploratory Learning Environments for the Classroom. In R. Luckin, S. Puntambekar, P. Goodyear, B. L. Grabowski, J. Underwood, N. Winters
(Eds.), Handbook of Design in Educational Technology (pp. 80-92). Routledge.
66
67. Classroom Dynamics
Gutierrez-Santos, S., Mavrikis, M., Geraniou E., Poulovassilis, A. (2012). Usage Scenarios and Evaluation of Teacher Assistance Tools for Exploratory Learning Environments (Under review)
Available at http://www.dcs.bbk.ac.uk/research/techreps/2012/bbkcs-12-02.pdf
67
70. S. Gutierrez-Santos; M. Mavrikis; E. Geraniou; A. Poulovassilis, "Similarity-based Grouping to Support Teachers on Collaborative
Activities in an Exploratory Mathematical Microworld," in IEEE Transactions on Emerging Topics in Computing , in press
Grouping
70
71. Common desired ‘superpowers’
Holstein, K., McLaren, B. M., & Aleven, V. (2017). Intelligent tutors as
teachers’ aides: Exploring teacher needs for real-time analytics in
blended classrooms. 7th International Conference on Learning Analytics
and Knowledge, Vancouver, Canada, March 13-17, 2017.
• students’ thought processes
• which students are really “stuck”
• which students are “almost there”, just need a nudge
• clone themselves
• have “eyes in the back of my head”
• know whether a student is actually trying
71
72. Emerging technology
• Wearable support tools
• Cross physical – digital
• Multimodal LA
• OLM for students
• Configurable summaries
Holstein et al (2017) LAK 2017
72
@mavrikis#VJornadasIHC
75. Summary
• AI and LA (perceptions) are changing rapidly
• Integration encourages adoption
• Focus on:
• Delegating teacher responsibility
• Actionable insights
• Context and user needs
7575
@mavrikis#VJornadasIHC
82. Rose Luckin Mutlu Cukurova,
Nikol Rummel
Sokratis KarkalasKaska Poryaska-Pomsta
FUNDERS
&
PROJECTS
Mina Vasalou
Beate Grawemeyer
Sergio Gutiérrez-Santos
Wayne Holmes
@mavrikis #VJornadasIHC @uclknowledgelab
82
83. Grawemeyer, B., Mavrikis, M., Holmes, W. et al. Affective learning: improving engagement and enhancing learning
with affect-aware feedback. UMUAI (2017) 27: 119. doi:10.1007/s11257-017-9188-z http://bit.ly/affect-italk2learn
84. Grawemeyer, B., Mavrikis, M., Holmes, W. et al. Affective learning: improving engagement and enhancing learning
with affect-aware feedback. UMUAI (2017) 27: 119. doi:10.1007/s11257-017-9188-z http://bit.ly/affect-italk2learn
85. • Talk aloud
• “Remember to talk aloud, and tell us what are you thinking”
• “What is the task asking you to do?”
• “Please think aloud, what are your thoughts or feelings?”
• Affect boosts
• “It may be hard, but keep trying”
• “If you find this easy, check your work and change the task”
• Problem solving
• “What do you need to do now, to complete the fraction?”
• Instructive feedback
• “You can’t add fractions with different denominators”
• Reflection
• “What did you learn from this task?”
• “What do you notice about the two fractions?”
Feedback types
85
86. Feedback framework
Holmes W., Mavrikis M., Hansen A., Grawemeyer B. (2015) Purpose and Level of Feedback in an Exploratory Learning
Environment for Fractions. In: Conati C., Heffernan N., Mitrovic A., Verdejo M. (eds) Artificial Intelligence in Education. AIED 2015.
Lecture Notes in Computer Science, vol 9112. Springer.
86
87. Pedagogic strategies for student support
• Supporting processes of exploration
• Supporting students to set and work towards explicit goals.
• Directing students’ attention.
• Helping students organise their working environment.
• Provoking cognitive conflict.
• Encouraging alternative solutions.
• Supporting reflection
• Promoting motivation
• Supporting collaboration
Mavrikis, M et al. (2008) "Revisiting pedagogic strategies for supporting students’ learning in mathematical microworlds."
Proceedings of the International Workshop on Intelligent Support for Exploratory Environments at EC-TEL. 87
Notes de l'éditeur
UCL Education and Technology Masters http://bit.ly/ucl-edtech19
Editor of British Journal of Educational Technology http://bit.ly/bjedtech
Discuss cases where you’ve seen (or can imagine) AI applied in Education
Terminology clarifications. Not talking about VR/AR (though can be enabled by AI). Or teaching of AI (though that’s important especially these days)
While anything is possible, not everything is the right thing we should be spending our time
Popular AI
So AI is becoming more and more popular through movies and the same applies for data analysis.
Movies like ex machine and transcendence may be actually damaging the perceiption of field with their technocentric portayal of AI.
When I talk to teachers about AI in education they usually say that we are building Minory Report to predict students failure but I think we aare more closer to Money Ball that revolves around a coach using data and computer analytics to judge and acquire players rather than on mere biases.
Showing that is more about asking the right questions than just having the data.
A vision of AIED
1992 - cyberpunk dystopian victorial times (post AI)
Primer is designed to react to the owners environment teach them what they need to know to survive.
Something like a Mentor (reference back to Telemachus)
Lack of agency
Critique useful but ignores research
I ‘ll make a case that it is important for these to come together
**led** by teaching and learning needs
The critics claim that the learner has little agency, is forced along a particular, even if individualised, path
This gives no opportunity to interact with the system about what is being learned.
They are often positioned as well as being possible to replace the teaacher
And being the low hanging fruit
But it is actually the argument that is the low hanging fruit as it weak
Still particular type of intelligence
People learn in other ways too.
So I mentioned one goal in the field is to design EEL
I use the term rather loosley
Whizz through some examples
3d logo
Some of you may know
Dynamic geometry
Domain language model areas
Game knows where to place children within the domain model – stores child’s performance in a learner model – adapts content and activity based on this performance
41
messages that approximate (NOT REPLACE!) human behaviour
To tap into students’ inner thoughts we introduced speech recognition technology
And what a better application for AI? A complex world of unstractued data to make sense of !
We address these challenges party with this architercture
06:00
06:00
perceived task difficulty classifier (PTDC) which uses prosodic cues in the student’s speech to predict the level
of challenge for the current student, and the output from speech recognition software which identifies words in the student’s speech.
06:00
his layer contains an affective state reasoner, implemented as a Bayesian network which draws on information from the learner model, in particular the student’s affective state, to decide what type of feedback should be provided to the student. The resulting feedback type is then stored in the learner model and provided to the feedback generation layer.
his layer contains an affective state reasoner, implemented as a Bayesian network which draws on information from the learner model, in particular the student’s affective state, to decide what type of feedback should be provided to the student. The resulting feedback type is then stored in the learner model and provided to the feedback generation layer.
I will not go into details here. But
25-30 minus here
Max 30 minutes here
So all nice but limitations !
AI Singularity is not close
- We know there are places where students will get stuck in new ways where AI fails
- And when that happens we know that AI will be perceived again as Black Box
62
But this is where my case for better integration of AI and LA comes
Most often than not the case is as in this cartoon that helps us remember that
67
68
And ok we may not be able to give them cloining but we can simualate eyes in back or whether students are
So I think so far we are only scratching the surface and some of the work that is emerging is quite powerful
And we designed different types of feedback that could be applied ranging from talk aloud to remind them to talk to targeting their affect to reflective prompts