3rd Workshop On Personalization Approaches For Learning .

2y ago
12 Views
2 Downloads
1.02 MB
32 Pages
Last View : 1m ago
Last Download : 3m ago
Upload by : Philip Renner
Transcription

3rd Workshop on Personalization Approaches forLearning Environments (PALE 2013)PrefaceMilos Kravcik1, Olga C. Santos2, Jesus G. Boticario2, Diana Pérez-Marín31RWTH University Aachen, Germanykravcik@dbis.rwth-aachen.de2aDeNu Research Group, Artificial Intelligence Department,Computer Science School, UNED, Spainocsantos@dia.uned.es – jgb@dia.uned.eshttp://adenu.ia.uned.es/3Laboratory of Information Technologies in Education (LITE).Universidad Rey Juan Carlos, Spaindiana.perez@urjc.esAbstract. Personalization approaches in learning environments can be addressed from different perspectives and also in various educational settings, including formal, informal, workplace, lifelong, mobile, contextualized, and selfregulated learning. PALE workshop offers an opportunity to present and discussa wide spectrum of issues and solutions, such as pedagogic conversationalagents, personal learning environments, and learner modeling.1IntroductionThe 3rd International Workshop on Personalization Approaches in Learning Environments (PALE) 1 takes place on June 10th, 2013 and is held in conjunction with the 21thconference on User Modeling, Adaptation, and Personalization (UMAP 2013). Thetopic can be addressed from different and complementary perspectives. PALE workshop aims to offer a fruitful crossroad where interrelated issues can be contrasted,such as pedagogic conversational agents, responsive open learning environments, andlearner modeling. The benefits of the personalization and adaptation of computerapplications have been widely reported both in e-learning (the use of electronic mediato teach, assess, or otherwise support learning) and b-learning (to combine traditionalface-to-face instruction with electronic media - blended 013/PALE 20131

Previous PALE workshops (both at UMAP 2011 and UMAP 2012) have shownseveral important issues in this field, such as behavior and embodiment of pedagogicagents, suitable support of self-regulated learning, appropriate balance between learner control and expert guidance, design of personal learning environments, contextualrecommendations at various levels of the learning process, predicting student outcomes from unstructured data, modeling affective state and learner motivation, andusing sensors to understand student behavior and tracking affective states of learners,harmonization of educational and technological standards, processing big data forlearning purposes, predicting student outcomes, adaptive learning assessment, andevaluation of personalized learning solutions. This points at individualization of learning as still a major challenge in education where rapid technological developmentbrings new opportunities how to address it. A lot of data can be collected in the educational process, but we need to find ways how to use it reasonably and to developuseful services in order to make the learning process more effective and efficient.Novel personalized services and environments are needed especially in lifelong andworkplace educational settings, in order to support informal, self-regulated, mobile,and contextualized learning scenarios. A big challenge is also adaptation consideringboth long-term objectives and short-term dynamically changing preferences of learners. Here open and inspectable learner models play an important role. In the case ofpedagogic conversational agents personalization is fostered by the use of adapteddialogues to the specific needs and level of knowledge of each student.In order to foster the sharing of knowledge and ideas to research on these issues,PALE format moves away from the classic 'mini-conferences' approach and followsthe Learning Cafe methodology to promote discussions on open issues regarding personalization in learning environments. This means that participants attending theworkshop benefit both from interactive presentations and constructive work.2Workshop themesThe higher-level research question addressed in the workshop is: “What are suitableapproaches to personalize learning environments?” It is considered in various contexts of interactive, personal, and inclusive learning environments. The topics of theworkshop included (but not limited to) the following: Motivation, benefits, and issues of personalization in learning environments Approaches for personalization of inclusive, personal and interactive learning environments Successful methods and techniques for personalization of learning environments Results and metrics in personalized learning environments Social and educational issues in personalized learning environments Use of pedagogic conversational agents Affective computing in personalized learning environments Ambient intelligence in personalized learning environments User and context awareness in personalized learning environmentsPALE 20132

3ContributionsA blind peer-reviewed process by three reviewers per paper with expertise in the areawas carried out to select the contributions for the workshop. As a result, 4 submissions were accepted, which report designing approaches, evaluation methods andopen issues for eliciting the recommendation support to personalize learning environments.Arevalillo-Herráez et al. [1] discuss what is needed to design an experiment forcapturing relevant information from an ITS to improve the learner’s competence insolving algebraic word problems considering learners’ emotional and mental states.To enrich learner’s experience with affective support both action logs to record user’sinteraction with the system, which can be used to discover important information thathelp instructional designers to improve the ITS performance, and emotional information gathered from external sources, which reflect affective or mental states, can beused.Labaj and Bieliková [2] propose a conversational evaluation approach be usedwithin ALEF adaptive learning framework that tracks the user attention and uses thatinformation to ask the evaluation questions at the appropriate time and right when theuser is working with the part in question (or just finished working with it). This approach aimed to get higher cooperation from the user providing more feedback thanwhen we would ask them randomly.Koch et al. [3] are researching, developing, and testing technologies to instrumentclassrooms, collect human signal data, and derive meaning that can lead to understandtheir relation with the education performance. In particular, they have developed aninterface to capture human signals in learning environment, integrated into innovativeanalytic models to extract meaning from these data and have implemented a proof-ofconcept experiment to detect variations of attention deficit hyperactivity disorderbased on level of attentiveness, activity and task performance.Manjarrés-Riesco et al. [4] discuss open issues which arise when eliciting personalized affective recommendations for distance learning scenarios, such as scarce reported experiences on affective support scenarios, ii) affective needs, iii) difficulties ofaffective communication in virtual learning communities, iv) reduced scope of theaffective support provided in current approaches, and v) lack of resources for educators to provide affective support. These issues were identified in the course of applying TORMES user centered engineering approach to involve relevant stakeholders(i.e. educators) in an affective recommendation elicitation process.AcknowledgementsPALE chairs would like to thank the authors for their submissions and the UMAPworkshop chairs for their advice and guidance during the PALE workshop organization. Moreover, we also would like to thank the following members of the ProgramCommittee for their reviews: Miguel Arevalillo, Maria Bielikova, Zoraida Callejas,Cristina Conati, Sabine Graf, David Griol, Judy Kay, Kinshuk, Ralf Klamma, TobiasPALE 20133

Ley, Ramón López-Cózar, Noboru Matsuda, Beatriz Mencía, Alexander Nussbaumer,Alexandros Paramythis, Dimitris Spiliotopoulos, Carsten Ullrich, and MartinWolpers. The organization of the PALE workshop relates and has been partially supported by the following projects: ROLE (FP7 IST-231396), Learning Layers (FP7318209) funded by the 7FP of the European Commission, and MAMIPEC (TIN201129221-C03-01) funded by the Spanish Ministry of Economy and Competence.References1. Arevalillo-Herráez, M., Moreno-Picot, S., Arnau, D., Moreno-Clari, P., Boticario, J.G.,Santos, O.C., Cabestrero, R., Quirós, P., Salmeron-Majadas, S., Manjarrés-Riesco, A.,Saneiro. Towards Enriching an ITS with Affective Support. In proceedings of the 3ndWorkshop on Personalization Approaches for Learning Environments (PALE 2013).Kravcik, M., Santos, O.C., Boticario, J.G. and Pérez-Marín, D. (Eds.). 21th conference onUser Modeling, Adaptation, and Personalization (UMAP 2013), 2012, p. 5-13.2. Labaj, M. Bieliková, M. Conversational Evaluation of Personalized Solutions for AdaptiveEducational Systems. In proceedings of the 3nd Workshop on Personalization Approachesfor Learning Environments (PALE 2013). Kravcik, M., Santos, O.C., Boticario, J.G. andPérez-Marín, D. (Eds.). 21th conference on User Modeling, Adaptation, and Personalization (UMAP 2013), 2012, p. 14-19.3. Koch, F., Ito, M., da Silva, A.B.M., Borger, S., Nogima, J. Exploiting Human Signals inLearning Environment as an Alternative to Evaluate Education Performance. In proceedings of the 3nd Workshop on Personalization Approaches for Learning Environments(PALE 2013). Kravcik, M., Santos, O.C., Boticario, J.G. and Pérez-Marín, D. (Eds.). 21thconference on User Modeling, Adaptation, and Personalization (UMAP 2013), 2012, p.20-25.4. Manjarrés-Riesco, A., Santos, o.C., Boticario, J.G., Saneiro, M. Open Issues in Educational Affective Recommendations for Distance Learning Scenarios. In proceedings of the 3ndWorkshop on Personalization Approaches for Learning Environments (PALE 2013).Kravcik, M., Santos, O.C., Boticario, J.G. and Pérez-Marín, D. (Eds.). 21th conference onUser Modeling, Adaptation, and Personalization (UMAP 2013), 2012, p. 26-33.PALE 20134

Towards Enriching an ITS with Affective SupportMiguel Arevalillo-Herráez1, Salvador Moreno-Picot1, David Arnau2, Paloma MorenoClari1, Jesus G. Boticario3, Olga C. Santos3, Raúl Cabestrero4, Pilar Quirós4, SergioSalmeron-Majadas3, Ángeles Manjarrés-Riesco3, Mar Saneiro31Department of Computer Science, University of Valencia, Spainmiguel.arevalillo@uv.es, salvador.moreno@uv.es, paloma.moreno@uv.es2Department of Mathematics Education, University of Valencia, Spaindavid.arnau@uv.es3aDeNu Research Group. Artificial Intelligence Dept. Computer Science School. UNED, Spainjgb@dia.uned.es, ocsantos@dia.uned.es, ssalmeron@bec.uned.es,amanja@dia.uned.es, marsaneiro@dia.uned.es4Department of Basic Psychology, Faculty of Psychology, UNED, Spainrcabestrero@psi.uned.es, pquiros@psi.uned.esAbstract. Recent progress in affective computing is having an important impacton the development of Intelligent Tutoring Systems (ITS). Many ITS use actionlogs to record user’s interaction with the system, such as to discover importantinformation that help instructional designers to improve the ITS performance.However, finer grain interaction data as well as emotional information gatheredfrom external sources is required to determine affective or mental states that canbe used to enrich learner’s experience with affective support. In this paper, wediscuss what is needed to design an experiment for capturing relevant information from an ITS to improve the learner’s competence in solving algebraicword problems considering learners’ emotional and mental states.Keywords: Affective computing, ITS, Multimodal emotions detection.1IntroductionUser’s affective state features a strong relationship with the cognitive process [1-4]. Inthe MAMIPEC and MARES projects we aim at exploring potential applications ofaffective computing in the context of accessible and personalized learning systems.To this end, we consider a user context that includes a wide range of appliances anddevices to enrich the user’s interaction. To study possible ways to detect user’s emotions in a learning context, a number of experiments focused on emotional data gathering have been carried out. A total of 92 subjects with different profiles and backgrounds, including people with functional diversity [5], were asked to solve a collection of mathematical exercises through dotLRN Learning Management System (LMS)while emotional information was gathered both from sensors and questionnaires.In order to further understand the learning implications of affective states, identifypossible applications of affect detection in tutoring systems, and reinforce some of thePALE 20135

conclusions drawn from the above study, we are currently following two researchdirections: 1) investigating potential applications of affective computing to improvean ITS developed in the context of the MARES project [6, 7]; 2) extending thedotLRN open source LMS and related software modules to include the required adaptive affective support through affective educational oriented recommendations [8].This paper describes some of the actions adopted by both research groups to improve the existing ITS and endow it with adaptive and affective support through recommendations. This ITS is deployed as a standalone application that provides tutoring features on a mathematical topic. In particular, the application aims at improvingthe learner’s competence in solving algebraic word problems. The algebra domain hasbeen chosen because of the many possibilities that it offers, regarding potential responses to specific mental states. Next, the ITS is described. After that, we discusshow to enrich the ITS with affective information based on the analysis of results carried out to date on the aforementioned experiments.2ITS description and position within the state of the artThe ITS emulates the behavior of a human tutor by tracking the current resolutionpath that the student is following, and adapts feedback accordingly. To this end, expert knowledge on the structure of word problems is codified by using hypergraphsthat represent the relations between quantities in the different analytical readings associated with each problem [6, 7]. The system is able to provide feedback and hintson demand. In both cases, the most likely analytical reading is computed and used toadapt the system response, which is given in natural language.Fig. 1. A screenshot of the ITS in tutoring modeFig. 1 shows a screenshot of the system in tutoring mode. The panel on the lefthand side is used to define quantities, either by using a letter or as a function of otherPALE 20136

quantities that have already been defined. In the figure, the student has already usedletters x and y to designate the ages of Mike and his father, respectively; and is currently defining Mike’s age 4 years ago as x-4. The panel on the right hand size is usedto build equations that relate several existing quantities. To encourage a systematicproblem solving approach, calculator-like components are used in both cases. Thesecontain the basic operators and one button per quantity already defined. The component used to build equations includes an additional button for the equals sign. In thisway, quantities need to be defined before they are used to either define another quantity or set an equation. The question mark button at the right-bottom corner of thescreen is used to request a hint. If this button is pressed, a hint is displayed on a floating window. This window is also used to provide feedback to incorrect actions. InFig.1, a sample help box is also shown on top of the main application window.The ITS has been designed so that action logs are dynamically produced as the userinteracts with the system. Student actions are written to a file in natural language. Fig.2 shows an example of the output generated. In this file, it can be observed that afterdefining the two letters, the student requested a hint. Again, the student felt unable tocarry out the recommended action and asked for further help. The system reacted bygiving further details on the first action suggested. Still, the student did not knowwhat to do and abandoned the application without finishing the resolution. Apart fromother obvious uses of such visual information (e.g. files can be inspected to study thestudent’s performance in detail), we are currently working on applying machine learning algorithms to the logs in order to draw relevant conclusions regarding situationsthat may demotivate the learner and cause abandonments.NEW PROBLEM LOADED: AgesSTATEMENT: Mike's father is 3 times as old as Mike. 4 years ago, hewas 4 times older. How old is Mike?USER ACTION: DEFINING LETTER.- x to represent Mike's current ageSYSTEM ACTION: ACCEPTEDUSER ACTION: DEFINING LETTER.- y to represent Mike's father ageSYSTEM ACTION: ACCEPTEDSYSTEM ACTION: HINT GIVEN.4 years ago, Mike was four years younger than todayYou may try to defineMike's age 4 years agoas a function of- 4- Mike's current age (x)SYSTEM ACTION: HINT GIVEN.4 years ago, Mike was four years younger than todayHenceMike's age 4 years ago Mike's current age less 4You may try to defineMike's age 4 years agoas x-4USER ACTION: EXITING WITHOUT FINISHINGFig. 2. An example of the high-level log produced by the applicationPALE 20137

Despite the possibilities offered by the high level information in the logs, finergrain interaction data may have a relatively higher importance to determine affect ormental states. For example, inactivity times, mouse movements or the time elapsedbetween clicks when defining an expression may provide important indicators relevant for the learning process. Combined with other (ideally non-invasive) sources ofinformation (webcams, eye tracking hardware), interaction data can be used to detectspecific emotional situations such as concentration, boredom, confusion or frustration[9-11]. In turn, this information can be directly used by the ITS to adapt commonresponses and/or handed to a recommender system to act in consequence [2].3Issues to consider for emotions detection in the ITSCurrently, we are trying to take advantage of the ITS tracking capabilities to enrichthe multimodal emotional data mining detection approach [12] by gathering moredetailed interaction and emotional information from the ITS and further exploit itsadaptive features. In a previous experience [5], which was planned by a multidisciplinary team that includes experts in different fields (mathematics education, psychology, programming, data mining, machine learning and modeling), participants had tosolve multiple choice mathematical exercises. Affective states were elicited at predetermined moments during the experience and gathered through several sources asfollows: i) physiological sensors (hearth rate, breath rate, temperature, galvanic skinresponse, blood pressure) in order to detect significant variations related to certainchanges in learner's affective state, ii) video recording (web cams, Kinect device, eyetracker) to find characteristic emotional meaningful facial gestures and attention foci,iii) interaction records (from mouse, keyboard and desktop) to identify behaviouralchanges, iv) standardized questionnaires (e.g. Big Five Inventory, General SelfEfficacy Scale, Positive and Negative Affect Schedule) to take into account certainaspects of participant’s personality and emotions and v) self-reports and scales (e.g.Self-Assessment Manikin) on their feelings and thoughts.In order to elicit several affective states, three groups of questions were prepared.The first one was easy if paper and pencil could be used, but participants were notallowed to do so. The next group of questions was limited in time, allowing less timethan needed in order to cause stress in the participants (they were told that time wassufficient enough to fulfill the task). In the third group o

Department of Basic Psychology , Faculty of Psychology, UNED Spain . rcabestrero@psi.uned.es, pquiros@psi.uned.es Abstract. Recent progress in affective computing is having an important impact on the development of Intelligent Tutoring Systems (ITS). Many ITS use action logs to record us

Related Documents:

Jamie Beckland, VP of Marketing at Janrain says, "I think we're in the trough of disillusionment when it comes to personalization." Without doubt, these are still early days for many forms of personalization -- in fact, a large number of respondents said it was too early to tell whether their personalization efforts were yielding results.

XSEDE HPC Monthly Workshop Schedule January 21 HPC Monthly Workshop: OpenMP February 19-20 HPC Monthly Workshop: Big Data March 3 HPC Monthly Workshop: OpenACC April 7-8 HPC Monthly Workshop: Big Data May 5-6 HPC Monthly Workshop: MPI June 2-5 Summer Boot Camp August 4-5 HPC Monthly Workshop: Big Data September 1-2 HPC Monthly Workshop: MPI October 6-7 HPC Monthly Workshop: Big Data

Key words: Semantic web, web service, Personalization, Automated reasoning. 1 Introduction Personalization in web-based applications [1], as a global tendency nowadays, aims at alleviating the burden of information overload by tailoring the information presented based on an individual and immediate user’s needs. Between the numerous examples

Quadrant for Multichannel Marketing Hubs”). This Magic Quadrant focuses solely on vendors that offer personalization engines as stand-alone solutions. Marketers typically use personalization engines in service of three primary use cases: 1. Digital commerce — to customize experiences across digital sal

Send an oversized postcard that is printed in color on both sides Use a padded mailer to pique the recipient’s interest Apply Personalization Liberally In direct mail, personalization is key. With personalization, or custo

COVID-19 crisis on eCommerce companies and their website personalization strategies. Read on to discover how the crisis has accelerated eCommerce marketers' ambitions for website personalization and their plans for 2021. 1. The COVID-19 pandemic saw eCommerce make 10 years' worth of growth in a 90-day period.

Human Behavior in the Personalization Economy Niklas Eder* 25 STAN. TECH. L. REV. 1 (2021) ABSTRACT Personalization has long been a feature of online services, shaping targeted advertisement and online manipulation. Corporations now seek to exploit the enormous economic potential of personalization beyond the confinements of the online space.

features on travel websites. Previous research has consistently utilized either e-personalization or privacy concern as marketing tools. Our results show a moderate and direct effect of e-personalization and privacy on consumer's attitudes toward a company Website, which in turn has a strong effect on purchase intention, and also the .