Schatten, C., Janning, R. & Schmidt-Thieme, L. (2015). Vygotsky Based Sequencing Without Domain Information: a Matrix Factorization Approach. Extended paper from CSEDU 2014, Barcelona.
Sequencing contents, like tasks, hints, and feedbacks, is an open issue for Intelligent Tutoring Systems. The common approach is based on domain analysis by experts, who characterize each content with skills involved and a difficulty level. In addition, Machine Learning based sequencers require a specific dataset collection to create users’ models and a sequencing policy, which needs to be tested online with strong ethical requirements and a high number of users. In this paper we design a simulated learning environment with customizable scenarios. We also show that a performance prediction method can be used to crate offline fully personalized students’ models and sequence contents without domain engineering/authoring effort. The performance prediction method is enhanced by a score-based policy inspired by Vygotsky’s concept of Zone of Proximal Development and shows promising results compared to curriculum based policies in the designed simulated environment.
Janning, R., Schatten, C., Schmidt-Thieme, L. (2015). Improving Automatic Affect Recognition on Low-Level Speech Features in Intelligent Tutoring System. 10th European Conference on Technology Enhanced Learning. EC-TEL 2015, Toledo, Spain.
Currently, a lot of research in the field of intelligent tutoring systems is concerned with recognising student’s emotions and affects. The recognition is done by extracting features from information sources like speech, typing and mouse clicking behaviour or physiological sensors. According to the state-of-the-art support vector machines are the best performing classification models for those kinds of features. However, single classification models often do not deliver the best possible performance. Hence, we propose an approach for further improving the affect recognition performance, which is based on ideas from ensemble approaches and feature selection methods. The approach is proven by experiments on low-level speech features extracted from data which was collected in a study with German students solving mathematical tasks. In these experiments the proposed approach reached on average an affect recognition performance improvement of about 59% in comparison to a single SVM.
Hansen, A. & Mavrikis, M. (2015). Learning mathematics from multiple representations: two design principles. Faro, Portugal, ICTMT12, 2015.
This paper describes two design principles for designing mathematics tasks using technology. These are: The parallel instantiations principle. Presenting students with a large number of nonprototypical instantiations simultaneously and non-transiently perturbs their thinking and supports thinking-in-change. The discriminating tools principle: Discriminating tools enable children to differentiate between the tools’ feedback to acquire knowledge from their use. Students will learn within the task rather than merely from the task. These principles were first developed in a study on 9-11 year olds’ geometric defining and later applied and amended in a larger study that encouraged 9-11 year olds’ conceptual understanding of fractions. The paper presents how the principles were applied within the two studies.
Hansen, A.,Mavrikis, M., Holmes, W. & Geraniou, E. (2015). Designing interactive representations for learning fraction equivalence. Faro, Portugal, ICTMT12, 2015.
This paper describes a study that investigated the use of a tool in an exploratory learning environment (ELE) designed to support students’ understanding of equivalent fractions. The study, part of a larger project, involved 67 9-11 year old students in England. It addressed the question: How does a partitioning tool support students’ conceptual understanding of equivalence? Data were collected through observations, students’ written work in an equivalence task, and a written self-reflection on learning at the end of their time in the ELE. Results showed that using the partitioning tool with an area representation was instrumental in challenging some students’ preconceived ideas about equivalent fractions and that the students were able to develop situated abstractions about fraction equivalence.
Grawemeyer, B., Mavrikis, M., Holmes, W., Hansen, A., Loibl, K., Gutiérrez Santos, S. (2015). The Impact of Feedback on Students’ Affective States. International Workshop on Affect, Meta-Affect, Data and Learning. Madrid, Spain, AMADL, 2015.
Affective states play a significant role in students’ learning behaviour. Positive affective states can enhance learning, while negative affective states can inhibit it. This paper describes a Wizard-of-Oz study that investigates the impact of different types of feedback on students’ affective states. Our results indicate the importance of providing feedback matched carefully to the affective state of the students in order to help them transition into more positive states. For example when students were confused affect boosts and specific instructive feedback seem to be effective in helping students to be in flow again. We discuss this and other ways to adapt the feedback, together with implications for the development of our system and the field in general.
Voß, L., Schatten, C., Mazziotti, C., Schmidt-Thieme, L. (2015). A Transfer Learning approach for applying Matrix Factorization to small ITS datasets. 8th International Conference on Educational Data Mining, Madrid, Spain, EDM 2015.
Machine Learning methods for Performance Prediction in Intelligent Tutoring Systems (ITS) have proven their efficacy; specic methods, e.g. Matrix Factorization (MF), however suer from the lack of available information about new tasks or new students. Data collections are extremely expensive and previous data of tasks under development are generally discarded. In this paper we show how this problem could be solved by applying Transfer Learning (TL), i.e. combining similar but not equal datasets to train Machine Learning models. In our case we obtain promising results by combining data collected of German fractions’ tasks (517 interactions, 88 students, 20 tasks) with their non-exact translation of a previously American US version (140 interactions, 14 students, 16 tasks). In order to do so we also analyze the performance of MF based predictors on smaller ITS’ samples evaluating their usefulness.
Janning, R., Schatten, C., Schmidt-Thieme, L. (2015). How to Aggregate Multimodal Features for Perceived Task Difficulty Recognition in Intelligent Tutoring Systems. Madrid, Spain, EDM 2015.
Currently, a lot of research in the field of intelligent tutoring systems is concerned with recognising student’s emotions and affects. The recognition is done by extracting features from information sources like speech, typing and mouse clicking behaviour or physiological sensors. Multimodal affect recognition approaches use several information sources. Those approaches usually focus on the recognition of emotions or affects but not on how to aggregate the multimodal features in the best way to reach the best recognition performance. In this work we propose an approach which combines methods from feature selection and ensemble learning for improving the performance of perceived task difficulty recognition.
Janning, R., Schatten, C., Schmidt-Thieme, L. (2015). Recognising Perceived Task Difficulty from Speech and Pause Histogram. 17th International Conference on Artificial Intelligence in Education, Madrid, Spain, AIED 2015.
Currently, a lot of research in the field of intelligent tutoring systems is concerned with recognising student’s emotions and affects. The recognition is done by extracting features from information sources like speech, typing and mouse clicking behaviour or physiological sensors. In former work we proposed some low-level speech features for perceived task difficulty recognition in intelligent tutoring systems. However, by extracting these features some information hidden in the speech input is loosed. Hence, in this paper we propose and investigate speech and pause histograms as features, which preserve some of the loosed information. The approach of using speech and pause histograms for perceived task difficulty recognition is evaluated by experiments on data collected in a study with German students solving mathematical tasks.
Grawemeyer, B., Gutierrez-Santos, S., Holmes, W., Mavrikis, M., Rummel, N., Mazziotti, C., Janning, R. (2015). Talk, Tutor, Explore, Learn: Intelligent Tutoring and Exploration for Robust Learning. Madrid, Spain, AIED 2015.
It is widely acknowledged that many children have difficulty learning fractions. Accordingly, we are developing the iTalk2Learn system, which aims to facilitate the robust learning of fractions by children in primary and early secondary education. The iTalk2Learn system integrates structured, practice-based tasks with exploratory, conceptually-oriented tasks, and intelligent affect-aware support. The system focuses on natural interaction via intuitive user interfaces, which includes speech recognition, and speech production.
Grawemeyer, B., Mavrikis, M., Holmes, W., Hansen, A., Loibl., K., Gutierrez-Santos, S.(2015). Affect Matters: Exploring the Impact of Feedback during Mathematical Tasks in an Exploratory Environment. Madrid, Spain, AIED 2015.
We describe a Wizard-of-Oz study that investigates the impact of different types of feedback on students’ affective states. Our results indicate the importance of matching carefully the affective state with appropriate feedback in order to help students transition into more positive states. For example when students were confused affect boosts and specific instruction seem to be effective in helping students to be in flow again. We discuss this and other effective ways to and implications for the development of our system and the field in general.
Grawemeyer, B., Mavrikis, M., Holmes, W., Gutierrez-Santos, S.(2015). Adapting Feedback Types According to Students’ Affective States. Madrid, Spain, AIED 2015.
Affective states play a significant role in students’ learning behaviour. Positive affective states can enhance learning, while negative ones can inhibit it. This paper describes the development of an affective state reasoner that is able to adapt the feedback type according to students’ affective states in order to evoke positive affective states and as such improve their learning experience. The reasoner relies on a dynamic Bayesian network trained with data gathered in a series of ecologically valid Wizard-of-Oz studies, where the effect of feedback on students’ affective states was investigated.
Mazziotti, C., Loibl, K., Rummel, N. (2015). Collaborative or Individual Learning Within Productive Failure. Does the Social Form of Learning Make a Difference? CSCL 2015, Gothenburg, Sweden.
Productive Failure (PF) – comprising initial problem solving and delayed instruction – has been proven effective for learning when compared to Direct Instruction (DI) in multiple studies with high school and university students. Although the problem-solving phase is usually implemented in a collaborative setting, the role of collaboration for the effectiveness of PF remains unclear. In two quasi-experimental studies we investigated whether collaborative as compared to individual learning in PF leads to more learning. We also tested whether the beneficial PF effect could be replicated with much younger students, namely 4th and 5th graders, than previous studies. Only our first study replicated the PF effect. While the first study did not reveal differences between collaborative and individual learning, in the second study individual learners even outperformed their collaborative counterparts in both PF and DI conditions. Against these findings, we discuss possible prerequisites for PF and propose an agenda for follow-up CSCL research.
Mazziotti, C., Holmes, W., Wiedmann, M., Loibl, K., Rummel, N., Mavrikis, M., Hansen, A., Grawemeyer, B. (2015). Robust Student Knowledge: Adapting to Individual Student Needs as They Explore the Concepts and Practice the Procedures of Fractions. AIED 2015, Madrid, Spain.
By offering students opportunities for self-determined exploration of concepts, exploratory learning environments (ELEs) facilitate students’ constructivist sense-making activities and support the acquisition of conceptual knowledge . However, in classroom settings, this is usually insufficient. In fact, students need to acquire robust knowledge – deep, connected and comprehensive knowledge that lasts over time, accelerates future learning and transfers easily to new situations [1, 2] – which does not only require conceptual knowledge (understanding ‘why’) but also requires procedural knowledge (knowing ‘how’) . Hence, apart from providing students with opportunities to acquire conceptual knowledge, we also need to offer them opportunities to acquire procedural knowledge. It is well known that by enabling students to practice problem-solving procedures step-by-step in structured learning environments, such as Intelligent Tutoring Systems (ITSs), students can acquire procedural knowledge. Accordingly, our approach is to combine ELEs with ITSs. However, this begs the question, how to optimally combine exploratory learning activities (ELA) and structured practice activities (SPA). Following a theory-driven approach, we have developed a pedagogical intervention model that selects and sequences activities that are appropriate for the individual learner. It is implemented as a rule-based system in a learning platform about fractions. The model’s decision-making process relies on the detection of each individual student’s level of challenge (i.e. whether they were under-, appropriately or over-challenged by the previous learning activity). Thus, our model flexibly adapts to each individual student’s needs and provides them with a unique sequence of learning activities. Our formative evaluation trials suggest that single components of the intervention model, such as the ELE, mostly achieve their aims. The interplay between the different components of the intervention model (i.e. the outcomes of sequencing and selecting exploratory and structured practice activities) is currently being evaluated.
Peer-reviewed Short Paper
Holmes, W., Mavrikis, M., Hansen, A. & Grawemeyer, B. (2015). Purpose and Level of Feedback in an Exploratory Learning Environment for Fractions. AIED 2015, Madrid, Spain.
Peer-reviewed Journal Paper
Hansen, A., Geraniou, E. & Mavrikis, M. (2015). Igniting Professional Development Through Cooperative Inquiry: Improving Teachers’ Technological Pedagogical Content Knowledge of Fractions. Journal for Mathematics Teacher Education (JMTE).
Grawemeyer, B., Holmes, W., Gutierrez-Santos, S., Hansen, A., Loibl, K., Mavrikis, M.(2015). Light-Bulb Moment? Towards Adaptive Presentation of Feedback Based on Students’ Affective State. Atlanta, IUI 2015.
Affective states play a significant role in students’ learning behaviour. Positive affective states can enhance learning, whilst negative affective states can inhibit it. This paper describes a Wizard-of-Oz study which investigates whether the way feedback is presented should change according to the affective state of a student, in order to encourage affect change if that state is negative. We presented ‘high-interruptive’ feedback in the form of pop-up windows in which messages were immediately viewable; or ‘low-interruptive’ feedback, a glowing light bulb which students needed to click in order to access the messages. Our results show that when students are confused or frustrated high-interruptive feedback is more effective, but when students are enjoying their activity, there is no difference. Based on the results, we present guidelines for adaptively tailoring the presentation of feedback based on students’ affective states when interacting with learning environments.
Schatten, C., Janning, R., Schmidt-Thieme, L. (2015). Integration and Evaluation of a Matrix Factorization Sequencer in Large Commercial ITS. Austin, Texas, AAAI15.
Correct evaluation of Machine Learning based sequencers require large data availability, large scale experiments and consideration of different evaluation measures. Such constraints make the construction of ad-hoc Intelligent Tutoring Systems (ITS) unfeasible and impose early integration in already existing ITS, which possesses a large amount of tasks to be sequenced. However, such systems were not designed to be combined with Machine Learning methods and require several adjustments. As a consequence more than a half of the components based on recommender technology are never evaluated with an online experiment. In this paper we show how we adapted a Matrix Factorization based performance predictor and a score based policy for task sequencing to be integrated in a commercial ITS with over 2000 tasks on 20 topics. We evaluated the experiment under different perspectives in comparison with the ITS sequencer designed by experts over the years. As a result we achieve same post-test results and outperform the current sequencer in the perceived experience questionnaire with almost no curriculum authoring effort. We also showed that the sequencer possess a better user modeling, better adapting to the knowledge acquisition rate of the students.
Janning, R., Schatten, C., Schmidt-Thieme, L.(2014). An SVM Plait for Improving Affect Recognition in Intelligent Tutoring Systems. ICTAI 2014, Limassol.
Usually, in intelligent tutoring systems the task sequencing is done by means of expert and domain knowledge. In a former work we presented a new efficient task sequencer without using the expensive expert and domain knowledge. This task sequencer only uses former performances and decides about the next task according to Vygotsky’s Zone of Proximal Development, that is to neither bore nor frustrate the student. We aim to support this task sequencer by a further automatically to gain information, namely students affect recognized from his speech input. However, the collection of the data from children needed for training an affect recognizer in this field is challenging as it is costly and complex and one has to consider privacy issues carefully. These problems lead to small data sets and limited performances of classification methods. Hence, in this work we propose an approach for improving the affect recognition in intelligent tutoring systems, which uses a special structure of several support vector machines with different input feature vectors. Furthermore, we propose a new kind of features for this problem. Different experiments with two real data sets show, that our approach is able to improve the classification performance on average by 49% in comparison to using a single classifier.
Grawemeyer, B., Mavrikis, M., Gutierrez-Santos, B. and Hansen, A. (2014). Interventions During Student Multimodal Learning Activities: Which and Why? EDM 2014, London.
This paper describes a Wizard-of-Oz (WoZ) study which investigates the potential of Automatic Speech Recognition (ASR) together with an emotion detector able to classify emotions from speech to support young children in their exploration and reflection whilst working with interactive learning environments. We describe a unique ecologically valid WoZ study in a classroom. During the study the wizards provided support using a script, and followed an iterative methodology which limited their capacity to communicate, in order to simulate the real system we are developing. Our results indicate that there is an effect of emotions on the acceptance of feedback. Additionally, certain types of feedback are more effective than others for particular emotions.
Janning, R., Schatten, C. and Schmidt-Thieme, L. (2014). Feature Analysis for Affect Recognition Supporting Task Sequencing in Adaptive Intelligent Tutoring Systems. EC-TEL 2014, Graz.
Originally, the task sequencing in adaptive intelligent tutoring systems needs information gained from expert and domain knowledge as well as information about former performances. In a former work a new efficient task sequencer based on a performance prediction system was presented, which only needs former performance information but not the expensive expert and domain knowledge. This task sequencer uses the output of the performance prediction to sequence the tasks according to the theory of Vygotsky’s Zone of Proximal Development. The aim of this paper is to support this sequencer by a further automatically to gain information source, namely speech input from the students interacting with the tutoring system. The proposed approach extracts features from students speech data and applies to that features an automatic affect recognition method. The output of the affect recognition method indicates, if the last task was too easy, too hard or appropriate for the student. Hence, as according to Vygotsky’s theory the next task should not be too easy or too hard for the student to neither bore nor frustrate him, obviously the output of our proposed affect recognition is suitable to be used as an input for supporting a sequencer based on the theory of Vygotsky’s Zone of Proximal Development. Hence, in this work is (1) proposed a new approach for supporting task sequencing by affect recognition, (2) presented an analysis of appropriate features for affect recognition extracted from students speech input and (3) shown the suitability of the proposed features for affect recognition for supporting task sequencing in adaptive intelligent tutoring systems.
Presentation at European Conference on Technology Enhanced Learning (EC-TEL 2014)
Janning,R., Schatten, C. and Schmidt-Thieme. L. (2014) Local Feature Extractors Accelerating HNNP for Phoneme Recognition. KI 2014, Stuttgart.
Artificial neural networks are fast in the application phase but very slow in the training phase. On the other hand there are state-of- the-art approaches using neural networks, which are very efficient in image classification tasks, like the hybrid neural network plait (HNNP) approach for images from signal data stemming for instance from phonemes. We propose to accelerate HNNP for phoneme recognition by substituting the neural network with the highest computation costs, the convolutional neural network, within the HNNP by a preceding local feature extractor and a simpler and faster neural network. Hence, in this paper we propose appropriate feature extractors for this problem and investigate and compare the resulting computation costs as well as the classification performance. The results of our experiments show that HNNP with the best one of our proposed feature extractors in combination with a smaller neural network is more than two times faster than HNNP with the more complex convolutional neural network and delivers still a good classification performance.
Presentation at 37th German Conference on Artificial Intelligence (KI 2014)
Mazziotti, C., Loibl, K. and Rummel, N. (2014). Eine Quasi-experimentelle Studie zur Rolle kooperativen Lernens für die Effektivität des Productive Failure Ansatzes [Investigating the Impact of Collaborative Learning on the Effectiveness of Productive Failure]. DGPs 2014, Bochum.
Lernansätze, bei denen Schülerinnen und Schüler (SuS) in kleinen Gruppen versuchen eine für sie unbekannte Aufgabe selbstständig zu lösen ehe sie in einer zweiten Lernphase Instruktion zur kanonischen Lösung erhalten, haben sich besonders zur Förderung von Verständniswissen als effektiv erwiesen (Productive Failure, Kapur, 2012). Dies entspricht dem Vorteil konstruktiver gegenüber aktiven Lernaktivitäten in Chis Hypothese (2009). Die Rolle der Kooperation, in der Aufgabenbearbeitungsphase ist jedoch bislang noch unklar (Collins, 2012). Da beim kooperativen Lernen elaborative Prozesse (Teasley, 1995) und damit interaktive Lernaktivitäten (Chi, 2009) angeregt werden, sollte kooperativ-interaktives Lernen in der Aufgabenbearbeitungsphase zu mehr Verständniswissen führen als individuell-konstruktives Lernen. Eine quasi-experimentelle Studie (N= 55) untersucht diese Hypothese: In zwei PF-Bedingungen bearbeiteten SuS zunächst eigenständig eine Aufgabe. Die SuS arbeiteten entweder paarweise-interaktiv (PF-Koop) oder individuell-konstruktiv (PF-Ind). Danach erhielten sie Instruktion. In einer dritten Bedingungen (DI) erhielten SuS zuerst Instruktion und bearbeiteten dann aktiv eine Aufgabe. Im Posttest unterschieden sich die beiden PF-Bedingungen nicht signifikant voneinander (F[1,48] = 0.4, p = .84). Der fehlende Unterschied zwischen PF-Koop und PF-Ind könnte andeuten, dass auch während der Kooperationen überwiegend konstruktive Aktivitäten zum Tragen kamen. Aufgrund der hohen Varianz in der PF-Koop-Bedingung (PF-Koop: M=7,88, SD=5,43; PF-Ind: M=7,71, SD=3,29; DI: M=5,26, SD=3,91) kann zudem angenommen werden, dass die Interaktionsgüte zwischen den einzelnen Paaren stark variierte und dass diese den Effekt moderiert. Beide Annahmen werden in laufenden Videoanalysen überprüft, die auf der Konferenz vorgestellt werden. Unabhängig von der Kooperation repliziert die Studie den PF-Effekt, also die Überlegenheit konstruktiver gegenüber aktiven Lernaktivitäten: SuS beider PF-Bedingungen erwarben signifikant mehr Verständniswissen als SuS der DI-Bedingung (F[1,48]=4.6, p=.03).
Contribution in edited volume for researchers and teachers, Ruhr-Universität Bochum
Mazziotti, C., Rummel, N. (2014). Kooperatives Lernen: Eine entscheidende Komponente der Productive Failure Methode? [Collaborative learning: A crucial component of Productive Failure?]. Contribution in an anthology for researchers and teachers. Ruhr-University Bochum.
Die sogenannte Productive Failure (PF) Methode (Kapur 2012) gehört zu einer Gruppe von Lernansätzen, bei denen Schülerinnen und Schüler (SuS) zuerst eine Einstiegsaufgabe in Kleingruppen bearbeiten ohne diese vorher im Unterricht behandelt zu haben und im Anschluss daran das zur Bearbeitung der Einstiegsaufgabe notwendige Wissen vermittelt bekommen. In etlichen Studien konnte gezeigt werden, dass derartige Lernansätze insbesondere den Erwerb von Verständniswissen fördern. Die Rolle des kooperativen Lernens, also der Zusammenarbeit in den Kleingruppen, bei der Bearbeitung der Einstiegsaufgabe ist jedoch bislang nicht hinreichend erforscht. Diese bestehende Forschungslücke soll im Dissertationsprojekt bearbeitet werden. Die zugrundliegende Annahme ist, dass das kooperative Lernen in Kleingruppen SuS dazu anregt, sich gegenseitig ihre Lösungsideen zu erklären und diese zu diskutieren. Dadurch werden elaborative, verständnisfördernde Prozesse initiiert, die insbesondere für den Erwerb von Verständniswissen förderlich sind. Daher sollten SuS, die die Einstiegsaufgabe in Kleingruppen bearbeiten, mehr von der PF-Methode profitieren als SuS, die die Einstiegsaufgabe individuell bearbeiten. Diese Hypothese wird in einer ersten Studie überprüft. Weitere Studien sollen beleuchten, wie die Stärken von kooperativem und individuellem Lernen im Rahmen der PF-Methode optimal genutzt werden können.
Grawemeyer, B., Mavrikis, M., Hansen, A., Mazziotti, C. and Gutierrez-Santos, S. (2014). Employing Speech to Contribute to Modelling and Adapting to Students’ Affective States. EC-TEL 2014, Graz.
Affect plays a significant role in students’ learning behaviour. Positive affective states can enhance learning, while negative ones can inhibit it. This paper describes how we provide intelligent support in a learning platform based on affect states. We discuss two components: an affective state detector to perceive affective states in speech during interaction with the platform; and an affective state reasoner to provide support, which aims at aligning the learner’s personal goal with the learning task to evoke positive affective states for an enhanced learning experience.
Mazziotti, C., Loibl, K. and Rummel, N. (2014). Towards Combining Structured Practice and Exploratory Learning Activities to Foster Robust Knowledge. Paper presented at the EARLI SIG 6 & 7 Meeting (Instructional Design & Learning and Instruction with Computers), Rotterdam, The Netherlands.
Robust knowledge consists of two types of knowledge: conceptual and procedural knowledge. While procedural knowledge is acquired through repeated practice and deepening of problem-solving procedures, conceptual knowledge can be facilitated by providing students with exploratory learning activities and encouraging reflection. In order to facilitate robust knowledge students should, thus, be provided with structured practice and exploratory learning activities. We aim at providing a theoretical framework for combining both learning activities that answers two questions: Should students start with structured practice or with exploratory learning activities? And how should the different types of learning activities be sequenced? The theoretical framework can inform the development of learning environments that foster robust knowledge.
Mavrikis, M., Grawemeyer, B., Hansen, A. and Gutierrez-Santos, S. (2014). Exploring the Potential of Speech Recognition to Support Problem Solving and Reflection. Wizards go to school in the elementary maths classroom. EC-TEL 2014, Graz.
The work described in this paper investigates the potential of Automatic Speech Recognition (ASR) to support young children’s exploration and reflection as they are working with interactive learning environments. We describe a unique ecologically valid Wizard-of-Oz (WoZ) study in a classroom equipped with computers, two of which were set up to allow human facilitators (wizards) to listen to students thinking-aloud while having access to their interaction with the environment. The wizards provided support using a script and following an iterative methodology that limited on purpose their communication capacity in order to simulate the actual system. Our results indicate that the feedback received from the wizards did serve its function (i.e. it helped modify students’ behaviour in that they did think-aloud significantly more than in past interactions and rephrased their language to employ mathematical terminology). Additional results from student perception questionnaires show that overall students find the system suggestions helpful, not repetitive and understandable. Most also enjoy thinking aloud to the computer but, as expected, some find the feedback cognitively overloading, indicating that more work is needed on how to design the interaction tipping the balance towards facilitating post-task reflection.
Paper presented at European Conference on Technology Enhanced Learning (EC-TEL 2014)
Schatten, C., Wistuba, M., Schmidt-Thieme, L. and Gutiérrez-Santos, S. (2014). Minimal Invasive Integration of Learning Analytics Services in Intelligent Tutoring Systems. ICALT 2014, Athens.
A common problem when trying to apply data mining techniques to improve educational systems is the disconnection between those who have the expertise (e.g. universities) and those who have access to the data (e.g. small companies).
Bringing expertise into educational in-production systems is complicated because companies are reluctant to invest a lot of effort into integrating new technology that they do not fully trust, while the technology cannot prove its worth without access to real, valid data. In this paper, we explore the requirements that machine learning systems have to be applied to specific learning problems (sequencing and performance prediction), and then propose a minimally invasive protocol for sequencing (based on web services) to easily integrate Learning Analytics Services into e–learning systems.
Schatten, C. (2014). Matrix Factorization Feasibility for Sequencing and Adaptive Support in ITS. EDM 2014, London.
Performance prediction has the potential of ameliorate the student model of an Intelligent Tutoring System by predicting whether or not a student has mastered a specific set of skills. Recently, a simulated learning process has shown how performance prediction methods based on Matrix Factorization can be used for continuous score prediction and for sequencing content through a policy inspired by Vygotsky’s concept of the Zone of Proximal Development. In this paper, we discuss the feasibility of the approach analysing a commercial system dataset. We evaluate performances of the score predictor and feasibility of the Vygotsky policy for sequencing tasks and providing adaptive support.
Janning, R., Schatten, C. and Schmidt-Thieme, L. (2014). Multimodal Affect Recognition for Adaptive Intelligent Tutoring Systems. EDM 2014, London.
The performance prediction and task sequencing in traditional adaptive intelligent tutoring systems needs information gained from expert and domain knowledge. In a former work a new efficient task sequencer based on a performance prediction system was presented, which only needs former performance information but not the expensive expert and domain knowledge. The aim of this work is to support this approach by automatically gained multimodal input like for instance speech input from the students. The proposed approach extracts features from this multimodal input and applies to that features an automatic affect recognition method. The recognised affects shall finally be used to support the mentioned task sequencer and its performance prediction system. Consequently, in this paper is (1) proposed a new approach for supporting task sequencing and performance prediction in adaptive intelligent tutoring systems by affect recognition applied to multimodal input, (2) presented an analysis of appropriate features for affect recognition extracted from students speech input and shown the suitability of the proposed features for affect recognition for adaptive intelligent tutoring systems, and (3) presented a tool for data collection and labelling which helps to construct an appropriate data set for training the desired affect recognition approach.
Presentation at Workshop on Feedback from Multimodal Interactions in Learning Management Systems at 7th International Conference on Educational Data Mining (EDM 2014)
Mazziotti, C., Loibl, K. and Rummel, N. (2014). Does Collaboration Affect Learning in a Productive Failure Setting? ICLS 2014, Boulder, Colorado.
This paper was presented at the International Conference of the Learning Sciences as part of the symposium “Combining Generation and Expository Instruction to Prepare Students to Transfer Big Ideas Across School Topics” (Glogger, 2014). The paper presents a quasi-experimental study showing that productive failure (i.e., generating own solution approaches to a preparation problem prior to explicit instruction) in mathematics (more precisely fractions) is also beneficial for a new age group, namely, elementary school children. Additionally, we varied if students worked on the preparation problem in groups or individually. A simple comparison did not reveal differences in learning outcomes depending on the social setting. However, further analyses of video data will provide more and differentiated information about effective processes during collaborative generative activities.
Janning, R., Schatten, C. and Schmidt-Thieme, L. (2014). Automatic Subclasses Estimation for a Better Classification with HNNP. ISMIS 2014, Roskilde.
Although nowadays many artificial intelligence and especially machine learning research concerns big data, there are still a lot of real world problems for which only small and noisy data sets exist. An example for such a real word problem with very noisy and challenging data is Phoneme Recognition. Applying learning models to those data may not lead to desirable results. Hence, in a former work a hybrid neural network plait (HNNP) for improving the classification performance on those data was proposed. To address the high intraclass variance in the former investigated data manually estimated subclasses for the HNNP approach were used. In this work is investigated on the one hand the impact of using those subclasses instead of the main classes for HNNP and on the other hand an approach for an automatic subclasses estimation for HNNP to overcome the challenging, expensive and time consuming manual labeling. The results of the experiments with two different real data sets – like the phoneme data set TIMIT – show that using automatically estimated subclasses for HNNP delivers the best classification performance and outperforms also single state-of-the-art neural networks as well as ensemble methods.
Presentation at 21st International Symposium on Methodologies for Intelligent Systems (ISMIS 2014)
Janning, R. (2014). Feature Analysis for Affect Recognition Supporting Task Sequencing in Adaptive Intelligent Tutoring Systems. EC-TEL 2014, Graz.
iTalk2Learn is presenting at the European Conference on Technology Enhanced Learning (EC-TEL 2014) the paper ‘Feature Analysis for Affect Recognition Supporting Task Sequencing in Adaptive Intelligent Tutoring Systems’. The paper presents an analysis of features gained from speech input of students interacting with an intelligent tutoring system and shows that there are statistically significant feature combinations for recognizing students affects.
Hansen, A., Mavrikis, M. and Geraniou, E. (2014). Professional Development Through Cooperative Inquiry: Improving Teachers’ Technological Pedagogical Content Knowledge of Fractions.2014. Journal of Mathematics Teacher Education, 2014.
This paper describes our efforts to effectively involve teachers in the design process of the exploratory learning environment FractionsLab in a way that also provides a professional development opportunity. While our primary objective was children’s mathematical development and trialling the learning environment in the classroom, this study investigated the support for teachers’ technological pedagogical content knowledge as they participated in the iterative design and development through dedicated workshops that marked the initiation of a ‘co-operative inquiry’ process. We focus specifically on teachers’ views of the workshops as a professional development activity and the lessons learnt on the impact it had on teachers through the lens of the technological, pedagogical, content knowledge (TPACK) framework. Through a case study, the paper highlights the potential of supporting a teacher to become fully immersed in an inquiry about the use of FractionsLab in the classroom. The scientific significance of the study is its potential to add to the currently limited literature-base related to how dissemination and cooperative inquiry opportunities can act as continuous professional development activity. The scholarly significance relates specifically to teachers’ technological pedagogical content knowledge of fractions when it accompanies a specific software design.
Hansen, A. (ed) (2014) Children’s Errors in Mathematics: Understanding Common Misconceptions in Primary Schools. 3rd Edition. London: Learning Matters/SAGE.
The 3rd edition of this very popular title has been published. The book aims to be a practical guide to students’ common errors and misconceptions in mathematics to be used in planning mathematics lessons. It also supports a deeper understanding of the difficulties encountered in children’s mathematical development. In this third edition more than 60 new misconceptions were added and many of these were identified as a direct result of the work related to fractions undertaken in schools in the iTalk2Learn project.
Schatten, C. and Schmidt-Thieme, L. (2014). Adaptive Content Sequencing without Domain Information. CSEDU 2014, Barcelona.
In order to sequence contents in ITS, sequencers necessitate domain information and previous students scores. In this scenario several problems arise. For multiple-skill contents the definition of continuous skills and difficulties by humans becomes subjective. Moreover, adaptive sequencing requires on-line testing limited by strong ethical requirements and small users number. In this paper we show that a performance prediction method can be used to sequence contents without domain engineering/authoring effort, ameliorating sequencing over common domain informed strategies. Moreover, we discuss if a synthetic learning process can be modeled in a plausible way in order to facilitate preliminary testing with sequencing. In conclusion we show that Matrix Factorisation is able to deal with all the important actual problems of Intelligent Tutoring Systems: personalisation, multiple skill content modelling, and difficulty.
Mavrikis, M., Gutierrez-Santos, S., Geraniou, E. and Noss, R. (2013). Design requirements, student perception indicators and validation metrics for intelligent exploratory learning environments. 2013, London.
Dr Mavrikis (IOE) and Dr Gutierrez-Santos (BBK), together with colleagues from the London Knowledge Lab, have co-authored an article relevant to iTalk2Learn: ‘Design requirements, student perception indicators and validation metrics for intelligent exploratory learning environments’. The article focuses on the design and validation of intelligent exploratory environments. The requirements elicited and lessons learned are used for the design of task-dependent support and the Fractions Lab.
Mavrikis, M. (2013). Knowledge Elicitation Methods for Affect Modelling in Education. 2013, London.
Dr Mavrikis (IOE) has co-authored an article relevant to iTalk2Learn together with colleagues from LKL/IOE and universities in the USA: ‘Knowledge Elicitation Methods for Affect Modelling in Education’. The article provides a resource for methodological decision making in that it reviews current knowledge elicitation methods that are used to aid the study of learners’ affect and to inform the design of intelligent technologies for learning, discusses advantages and disadvantages of the specific methods as well as issues related to the interpretation of data that emerges as the result of their use.