Autonomy in English language learning: a case study

The main goal of this study was to analyse the development of a pedagogy for autonomy tuned to the current situation in which Higher Education finds itself, namely of implementation of the Bologna Process in Europe, and in particular at the Escola Superior Agrária of the Polytechnic School at Coimbra. For this purpose new course units were created, new descriptors written, new materials developed and new methodologies tried. A case study was undertaken to observe and analyse the changes the institution was undergoing and to assess the impact of the new methodologies, particularly regarding the development of student autonomy. The target population for this study were the first year students on the Biotechnology degree in the year 2007/2008, to whom a battery of tests were applied, namely proficiency tests, belief tests, metacognitive tests, as well as methods of ethnographic research. The study relied on the European Framework of Reference for Languages and the European Language Portfolio to establish recognised levels and competences, and employed the portfolio as its main tool for measuring student autonomy. This portfolio was simultaneously biographical and descriptive and was organised by the students in their own time. Great incentive was given to this independent work. The ethnographic component of the study was of significant importance being given to students learning histories and to the collection of students’ reflections. This component aimed in particular to allow students the opportunity to express their thoughts on the experience they underwent during 20 weeks. From the analysis of the data, a significant improvement in self-confidence and motivation for lifelong learning is observable in the development of multiple tasks. Regarding the students’ progression towards proficiency, this improvement is less clear, but the final results of educational success were generally of great relevance.

Avaliação do desempenho docente e supervisão pedagógica: um estudo de caso num agrupamento de escolas

A introdução de novas exigências na Avaliação do Desempenho Docente (ADD) em Portugal (Decretos-Lei 15/2007 e 75/2010; Decretos Regulamentares 2/2008 e 2/2010) tem gerado nas escolas e nos professores, desde 2008, situações organizacionais complexas, que urge analisar e compreender. As políticas educativas têm vindo a apontar para a necessidade de um maior investimento nos processos de avaliação de professores, dando relevância ao papel a desempenhar pela supervisão pedagógica nesse contexto. A supervisão da prática letiva passou a ser considerada procedimento fundamental, na medida em que se tornou difícil conceber uma avaliação dos docentes que não incluísse sessões de observação em sala de aula. Fruto destas novas políticas de avaliação de professores, a ADD apresentou-se, do ponto de vista normativo, com base em duas funções: sumativa e formativa. Estas funções exigem maiores investimentos, novas práticas e uma reflexão sobre o papel do professor ancorada numa perspetiva holística, transformadora e emancipatória do seu desenvolvimento profissional. O principal objetivo do nosso estudo consiste em compreender as perceções e práticas desenvolvidas pelos docentes face às alterações introduzidas no sistema da ADD no geral, e, em particular, o papel desempenhado pela supervisão pedagógica (SP) na avaliação entre pares, no segundo ciclo avaliativo (2009-2011). A investigação empírica tem por base um estudo de caso desenvolvido num Agrupamento de escolas do Distrito de Aveiro. A recolha de dados foi concretizada através da administração de um questionário a todos os docentes do Agrupamento e da condução de entrevistas a 26 informadores privilegiados do processo avaliativo. Os principais resultados do estudo de caso apontam para: i) a opinião desfavorável dos professores face ao atual modelo de ADD, dada a sua natureza burocrática, complexa, injusta e a sua escassa utilidade prática, preenchendo apenas requisitos legais e de controlo; ii) um processo de SP, concretamente, a observação de aulas, que não contribui para o desenvolvimento profissional dos professores e tem impacte negativo no trabalho docente, acentuando a artificialidade dos procedimentos e aumentando o clima de competição e de conflito; iii) a sensação de desconforto entre os intervenientes, fazendo emergir sentimentos de injustiça, insegurança e ansiedade que, em seu entender, constituem constrangimentos à melhoria das aprendizagens dos alunos e à qualidade da escola; iv) uma escassa implicação na melhoria das práticas futuras e do desenvolvimento profissional dos docentes. Estes resultados constituem um desafio colocado aos investigadores no sentido de identificarem alternativas de ação para poderem lidar com a complexidade inerente ao processo de avaliação na sua relação com a supervisão pedagógica, com vista à melhoria de desempenho das escolas e dos professores.

Techniques for batch reinforcement learning in robotics

This thesis addresses the Batch Reinforcement Learning methods in Robotics. This sub-class of Reinforcement Learning has shown promising results and has been the focus of recent research. Three contributions are proposed that aim to extend the state-of-art methods allowing for a faster and more stable learning process, such as required for learning in Robotics. The Q-learning update-rule is widely applied, since it allows to learn without the presence of a model of the environment. However, this update-rule is transition-based and does not take advantage of the underlying episodic structure of collected batch of interactions. The Q-Batch update-rule is proposed in this thesis, to process experiencies along the trajectories collected in the interaction phase. This allows a faster propagation of obtained rewards and penalties, resulting in faster and more robust learning. Non-parametric function approximations are explored, such as Gaussian Processes. This type of approximators allows to encode prior knowledge about the latent function, in the form of kernels, providing a higher level of exibility and accuracy. The application of Gaussian Processes in Batch Reinforcement Learning presented a higher performance in learning tasks than other function approximations used in the literature. Lastly, in order to extract more information from the experiences collected by the agent, model-learning techniques are incorporated to learn the system dynamics. In this way, it is possible to augment the set of collected experiences with experiences generated through planning using the learned models. Experiments were carried out mainly in simulation, with some tests carried out in a physical robotic platform. The obtained results show that the proposed approaches are able to outperform the classical Fitted Q Iteration.