Negative Wash-Back of Formative Assessment to Learning in Saudi Higher Education Context

This article reports the wash-back of formative assessment on what students learn, how they learn and the depth of their learning in Saudi higher education context. Previous research indicates that assessment methods affect different aspects of learning either positively or negatively depending on the nature of assessment tasks. Observations indicate a clear association between Saudi students’ learning and how their learning is assessed; so this research was needed to determine how exactly the correlation looked like—positive or negative. The data in this study were collected from Saudi undergraduates by employing a student survey and semi-structured interviews. The survey included Likert scale items of agreement regarding research assignments, quizzes and midterm examinations administered to 250 English-major students. To validate the survey results, sixteen students from different levels with GPA 3 and above were interviewed. The results showed that formative assessment narrowed down the scope of learning materials. The students mostly adopted surface level learning strategies to prepare for formative assessment tasks. Higher order thinking skills were not tested in any of the formative assessment methods. Therefore, it is suggested that assessments tasks should be subjected to thorough validation and moderation. Sound assessment practices should be put in place and practiced judiciously. To achieve these objectives, sustained institutional and departmental professional backing is a prerequisite.

Ontology Based Semantic Concurrent Activity Recognition

Activity recognition has a vital role in smart home operations. Major challenges in activity recognition are personalization, recognising parallel and interleave activities, erratic degree of dissimilar activities, identification of same object used in multiple activities, catering sensor noise caused by mal-interactions, dynamically determining the context of personalized activities and evolution of generic activity model for new activities. Moreover, object-sensor-based activity recognition by learning for complete activity pattern derived from a generic activity model in sequential and parallel activities may also be asserted as open research realms. A dynamic and generic framework named Ontology driven Semantic Activity Recognition (OSCAR) has been proposed to address the asserted challenges through hybrid of data driven techniques, temporal formalism and knowledge-driven techniques. An unlabelled sensor stream generated by inhabitant’s interactions has been accumulated into sensor repositories that is processed by OSCAR to recognise personalized activities performed in sequential or interleaved fashion. The major modules of OSCAR for activity recognition are sensor properties sequencer, semantic segmentor, personalized activity learner, spurious filter model and ontology evolution model. The spurious semantic segmentation produced by sensor noise or erratic behaviour is removed by Allen’s temporal formalism. Moreover, Tversky’s feature-based similarity has been used to remove the highly similar spurious activities produced as a result of mistaken interactions with wrong home objects. A comprehensive set of experiments has been carried out for evaluating the effectiveness of OSCAR over different metrics such as chi-square distribution, precision, recall and f-measure. In order to measure the performance of proposed technique covering all the possible actions/activities. A standard dataset, named CASAS, has been used for making a comparative analysis of different scenarios in activity recognition with state of the art work by Riboni and KCAR. In order to validate distinct research perspectives such as sensor noise, learning user specific actions; no dataset could comprehend these scenarios to the best of our knowledge. So, a dataset named Data Acquisition Methodology for Smart Homes (DAMSH) was developed while adhering to standard guidelines. The evaluation using stated metrics, over different datasets and comparative analysis with prevalent techniques assert OSCAR as a viable and superior solution. The efficacy of OSCAR is complemented by the distinctive features of dynamically learning personalized actions of inhabitants, boundary detection of activities, ontologies, identification and elimination spurious actions and seed knowledge evolution through ontologies.