Using AI to understand everyday learning on the Web
1 like362 views
1) The document discusses using artificial intelligence to understand informal learning that occurs on the web through people's everyday activities like searching online. 2) It describes several research projects aimed at detecting learning behaviors and predicting users' knowledge gains from analyzing patterns in their search histories, browsing activities, and other online traces. 3) The goal is to develop models that support learners in efficiently finding reliable information online and gauging their "learning to learn" skills, and applying these to specific online platforms commonly used for daily learning.
![08/11/18 6Stefan Dietze
Learning while searching the Web (“Search As Learning”)?
Challenges & results
Detecting coherent search missions?
Detecting learning throughout search?
detecting “informational” search missions (as
opposed to “transactional” or “navigational”
missions [Broder, 2002])
o Search mission detection with average F1
score 75% (experiments based on AOL query
logs, [CHIIR19])
How competent is the user? –
Predict/understand knowledge state of users in
absence of assessment data
How well does a user achieve his/her learning
goal/information need? - Predict knowledge gain
throughout search missions
o Correlation of user behavior (queries,
browsing, mouse traces, etc) with user
knowledge gain/state in various search tasks
[CHIIR18]
o Prediction of knowledge gain/state through
supervised models [SIGIR18, WWW19]](https://image.slidesharecdn.com/digitalenlightenmentforum2018sd-181109160326/85/Using-AI-to-understand-everyday-learning-on-the-Web-6-320.jpg)
Using AI to understand everyday learning on the Web
- 1. Using AI to understand everyday learning on the Web Prof. Dr. Stefan Dietze GESIS / Heinrich Heine University (Germany) 08.11.2018, Digital Enlightenment Forum, 8 November 2018, Brussels
- 2. Learning on the Web 08/11/18 2Stefan Dietze
- 3. Learning Resources on the Web? - LinkedUp Catalog Dataset Catalog/Registry http://data.linkededucation.org/linkedup/catalog/ “LinkedUp” (FP7 project): L3S, OU, OKFN, Elsevier, Exact Learning Solutions Publishing and curation of educational/learning resources according to Linked Data principles Largest collection of Linked Data about learning resources (approx. 50 datasets, 50 M resources) 08/11/18 3Stefan Dietze
- 4. 08/11/18 4Stefan Dietze Anything can be a learning resource (when used for „informal, „just in time“, micro-learning) The activity makes the difference (not the resource): i.e. how a resource is being used Challenges: o How to detect „learning“? o How to detect learning-specific notions such as „competence“, „learning performance“ etc? Learning Analytics in online/non-learning environments? o Activity streams (e.g. Twitter), o Social graphs (and their evolution), o Behavioural traces (mouse movements, keystrokes) o ... Analytics for everyday (online) learning? (as opposed to “education”) Figure courtesy of Mathieu d‘Aquin
- 5. 08/11/18 5Stefan Dietze Analytics for everyday (online) learning? (as opposed to “education”) AFEL: H2020 project (since 12/2015) aimed at understanding/supporting learning in social Web environments SALIENT: Leibniz project with various German research labs and educational organisations Learning efficiently on the Web = finding reliable and relevant information for a particular topic/learning need „Learning to learn“ = supporting learners/users to find information efficiently Figure courtesy of Mathieu d‘Aquin
- 6. 08/11/18 6Stefan Dietze Learning while searching the Web (“Search As Learning”)? Challenges & results Detecting coherent search missions? Detecting learning throughout search? detecting “informational” search missions (as opposed to “transactional” or “navigational” missions [Broder, 2002]) o Search mission detection with average F1 score 75% (experiments based on AOL query logs, [CHIIR19]) How competent is the user? – Predict/understand knowledge state of users in absence of assessment data How well does a user achieve his/her learning goal/information need? - Predict knowledge gain throughout search missions o Correlation of user behavior (queries, browsing, mouse traces, etc) with user knowledge gain/state in various search tasks [CHIIR18] o Prediction of knowledge gain/state through supervised models [SIGIR18, WWW19]
- 7. Predicting knowledge gain/state of user during search? 08/11/18 7Stefan Dietze Prediction through supervised machine learning models (after 10-fold cross-validation) KG prediction performance Feature importance (KG prediction) Ran Yu, Ujwal Gadiraju, Peter Holtz, Markus Rokicki, Philipp Kemkes and Stefan Dietze. Analyzing Knowledge Gain of Users in Informational Search Sessions on the Web. ACM SIGIR 2018. Key findings User knowledge gain / state can be predicted from user behavior during search missions In particular browsing behavior and queries of importance Ongoing work: investigating resource features (e.g. document complexity, analytic/emotional language tone, multimodality of resources) as additional signals Turning such models into actual applications => SALIENT & AFEL project
- 8. 08/11/18 8Stefan Dietze Outlook: supporting learning in online platforms Search as part of domain-specific and cross-domain online platforms (“bringing learning analytics to platforms where users learn on daily basis”) Examples (SALIENT & AFEL project) o Didacatalia – social online community of approx. 200.000 users o gesisDataSearch search of > 100.000 research datasets in the social sciences o Social Sciences Open Access Repository (SSOAR) – online archive of social sciences literature o TIB AV Portal – lecture videos and tutorials provided by the German National Library of Science & Technology http://GNOSS.com http://datasearch.gesis.org https://www.gesis.org/ssoar/home/ https://av.tib.eu/