Robots are destined to move beyond the caged factory floors towards domains where they will be interacting closely with humans. They will encounter highly varied environments, scenarios and user demands. As a result, programming robots after deployment will be an important requirement. To address this challenge, the field of Learning from Demonstration (LfD) emerged with the vision of programming robots through demonstrations of the desired behavior instead of explicit programming. The field of LfD within robotics has been around for more than 30 years and is still an actively researched field. However, very little research is done on the implications of having a non-robotics expert as a teacher. This thesis aims to bridge this gap by developing learning from demonstration algorithms and interaction paradigms that allow non-expert people to teach robots new skills. The first step of the thesis was to evaluate how non-expert teachers provide demonstrations to robots. Keyframe demonstrations are introduced to the field of LfD to help people teach skills to robots and compared with the traditional trajectory demonstrations. The utility of keyframes are validated by a series of experiments with more than 80 participants. Based on the experiments, a hybrid of trajectory and keyframe demonstrations are proposed to take advantage of both and a method was developed to learn from trajectories, keyframes and hybrid demonstrations in a unified way. A key insight from these user experiments was that teachers are goal oriented. They concentrated on achieving the goal of the demonstrated skills rather than providing good quality demonstrations. Based on this observation, this thesis introduces a method that can learn actions and goals from the same set of demonstrations. The action models are used to execute the skill and goal models to monitor this execution. A user study with eight participants and two skills showed that successful goal models can be learned from non- expert teacher data even if the resulting action models are not as successful. Following these results, this thesis further develops a self-improvement algorithm that uses the goal monitoring output to improve the action models, without further user input. This approach is validated with an expert user and two skills. Finally, this thesis builds an interactive LfD system that incorporates both goal learning and self-improvement and evaluates it with 12 naive users and three skills. The results suggests that teacher feedback during experiments increases skill execution and monitoring success. Moreover, non-expert data can be used as a seed to self-improvement to fix unsuccessful action models.

This report investigates the relationship between demonstrations and the built environment of the Forum Romanum. As one of the chief loci for the creation of public discourse in Rome, the Forum Romanum was a prime target for demonstrations. An in-depth evaluation of late Republican demonstrations within the Forum reveals how demonstrations sought to create alternative discourses. Late Republican demonstrators often incorporated the topography of the Forum into their demonstrations, either for strategic or symbolic reasons. Demonstrators were particularly concerned with the occupation of the Forum and restricting access to the speaker’s platforms. In doing so, demonstrations attempted to legitimate their own goals and objectives by equating them with the will of the people. The Augustan transformation of the Forum Romanum disrupted this established Republican topography of demonstration. Changes in the built environment limited the effectiveness of a demonstration’s ability to occupy the Forum. Entrances to the Forum were narrowed to impede the movement of demonstrators. Speaker’s platforms were insulated from the assembled crowd. A number of redundant measures, including surveillance and legal remedies, ensured that a new topography of demonstration did not form. These changes to the Forum Romanum participated in Augustus’s larger ideological program by prohibiting the creation of discourses opposed to the Augustan message. / text

This thesis analyses the method, purpose, and results of Aristotle's treatment of a select number of Presocratic and Platonic theses about the soul within the context of De Anima 1. Contrary to a prevalent interpretation of De Anima 1 which sees Aristotle's treatment of his predecessors' psychological views as dialectical, I argue that Aristotle treats his predecessors as having offered potentially viable hypotheses about the nature of the soul, and that these hypotheses are treated as attempts to explain the soul's basic powers. I further show that, in order to test the explanatory limits of these theories, Aristotle uses a version of the scientific method of inquiry advertised in the Prior and Posterior Analytics, which consists in setting out the basic psychological phenomena which psychology should explain, and then testing the extent to which his predecessors' definitions of the soul are able to accomplish this task. This thesis argues that, by demonstrating where his predecessors' first principles fall short, Aristotle is able to make theoretical progress towards establishing his own 'hylomorphic' first principle of soul, that soul is the 'first fulfilment of a natural instrumental body', and towards the idea that soul operates in the body of living beings as a formal, final, and efficient cause.

International Telemetering Conference Proceedings / October 17-20, 1994 / Town & Country Hotel and Conference Center, San Diego, California / A demonstration of near real-time performance assessment for the Program Executive Officer for Cruise Missiles Project and Unmanned Aerial Vehicles, Cruise Test Directorate, PEO(CU)-CT, was conducted between 22 March 1994 through 4 May 1994. The demonstration involved the temporary installation of a portable TOMAHAWK telemetry recording and telecommunications capability at the Air Force Development Test Center range at Eglin Air Force Base, Florida and a receiving telecommunications capability at the Naval Warfare Assessment Division (NWAD), Corona, California. The system was successfully used on 4 May 1994 to record TOMAHAWK missile telemetry data in real-time in support of Operational Test Launch (OTL)-163 and to transfer that data to the weapons system performance analysts at NWAD in near real-time. The one hour and three minutes of flight data was compressed in real-time as it was recorded, then, after completion of the flight, the data was transferred to NWAD in about 12 minutes using the switched 56 kbps network. Additional transfers using the Defense Commercial Telecommunications Network (DCTN) were also conducted. All transfers were secured using ethernet encryptors. The data was processed by both the NWAD telemetry ground station and the TOMAHAWK workstation complex. This paper quantifies the results and documents the lessons learned from this demonstration and proposes a standardized system design for possible implementation at TOMAHAWK test range sites in the future. A position is taken that for situations where the remote site (e.g. other range or data analysis site) does not exercise direct operational control over the test/host range, near real-time data relay solutions are not only as adequate, but in many cases are preferable to real-time solutions.

1 Web sites offer a huge amount of information. Often it is a page generated from data stored in databases. However, emphasis is placed on the display of information, but not on their machine processing. Part of the thesis is design and implementation of a prototype system to retrieve data from dynamically generated web using programming by demonstration technique. Such a system allows the user to show with mouse to the system how to proceed with gathering information from the website. Based on such a example, the system will derive a procedure to acquire information on similar sites. The implemented system is able to collect user relevant information from similar sites for example in form of a simple table suitable for further machine processing.

La question principale de cette thèse est la pertinence de l’utilisation d’humains virtuel pour l’enseignement de geste moteur. Une première étude explore la question de la faisabilité d’un apprentissage par l’imitation d’un humain virtuel. Pour ce faire, l’amélioration de performance de trois gestes de karaté par trois groupes s’entraînant dans différents environnements est évaluée. Une classe traditionnel de karaté, un enseignement par vidéo et un enseignement en réalitévirtuel sont comparé. Une seconde étude évalue l’influence d’une représentation de soi dans l’environnement virtuel sur la réalisation de la tâche d’apprentissage. Les participants à cette étude ont un retour visuel de leur action sous la forme d’un avatar cylindrique représenté dans un miroir de l’environnement virtuel. L’influence de cet avatar sur la tâche d’apprentissage est évalué de deux façon : 1) par l’évaluation externe de la performance 2) par le biais d’entretiens d’explicitation qui permettentd’obtenir le point de vue du participant. Ces deux études sont complétées par une troisième étude qui évalue un outil d’évaluation automatique de la performance. Cet outil est conçue dans le but d’éliminer les biais créés par un évaluateur humain. Un tel outil est nécessaire afin de pouvoir comparer les quatre environnements d’apprentissage présentés dans cette thèse ainsi que ceux qui seront créé dans le futur. En effet, seul un évaluateur automatique peut être parfaitement consistent sur toute la durée des études.En conclusion, ces études ont permit d’établir que l’apprentissage par l’imitation d’un humain virtuel est possible. En conséquence, l’utilisation d’humains virtuels pour l’enseignement de gestes moteurs est pertinente. Ces environnements peuvent être utiles à plusieurs fins. Ils peuvent servir d’environnements contrôles pour l’étude de l’apprentissage par imitation. Le protocole développé dans cette étude peut être utilisé dans une boucle itérative pour le développementde version subséquentes d’environnements virtuel ou d’accélérateurs d’entraînement virtuel. Ce protocole peut également être utilisé pour le développement de jeux vidéos actifs plus efficaces pour combattre la sédentarité et la crise mondiale d’obésité qui y est rattachée / The main question of that thesis is on the relevance of using virtual humans to teach complex motor skills. The first study explores the question of the feasibility of learning by imitation of a virtual human by comparing the improvement of the performance on three karate gestures for three groups, namely a traditional class, a video-based group and a virtual reality group. The second study investigates the influence on the learning task of having a self representation in the virtual environment. The participants have a feedback of their movements represented on a mirrored cylindrical gray avatar. The impact of that avatar on the learning task of the participants is assessed by two means. Performance evaluations are performed and give an external perspective on the learning. Evocation interview are also performed to get an insight of the learning task from the participants point of view. Finally, these two studies are completed by a third one investigating the possibility to have an automatic performance evaluator in order to reduce grading discrepancies generated by humans graders. Such a tool would be required to have an objective performance evaluation of all the participants in order to compare the four learning environments presented in that thesis and eventual further iteration of these environments. The conclusion our the studies presented in that thesis are that learning motor skills from the imitation of a virtual human is possible. Consequently, virtual learning environments for motorskills teaching are relevant. Furthermore, these environments can be used in various types of applications. They can be used as a study tool for standard and controlled investigation of teaching by demonstration. They can also be used in an engineering loop for the development of further learning environments and training accelerators. They also have a potential usage in the development of exergames in response to the international obesity crisis

[pt] Esta dissertação tem como objetivo compreender as Orientações Pedagógica e Psicopedagógica que possibilitaram a construção da experiência educativa da Escola Guatemala, instituição de caráter experimental, vinculada ao Departamento de Aperfeiçoamento do Magistério (DAM) do Centro Brasileiro de Pesquisas Educacionais (CBPE/INEP), com sede no Rio de Janeiro, à época, Distrito Federal. Por ser tratar de uma pesquisa no campo da História da Educação, elaborada sob a perspectiva da História Cultural, recorremos a entrevistas e à análise documental dos arquivos pessoais de Lúcia Marques Pinheiro e Terezinha Lins de Albuquerque, Supervisora Pedagógica do INEP e Coordenadora do Serviço de Orientação Psico-Pedagógica (SOPP), respectivamente. Partindo de uma interlocução permanente com Marcus Vinicius Cunha, foi possível apreender os tipos de orientação que os docentes desta escola receberam durantes os anos de 1955 a 1968, período compreendido como o final do movimento da Escola Nova no Brasil. Se, de um lado, a orientação pedagógica norteava a prática docente e, por sua vez, limitava as possibilidades de manifestação da subjetividade destes que ensinavam, do outro, a orientação psicopedagógica proporcionava a criação de um clima favorável e propício junto aos professores para que, assim, as propostas pedagógicas pudessem ser colocadas em prática. Deste modo, no primeiro capítulo buscamos caracterizar o contexto político e educacional que marca o período; no seguinte, descrevemos a maneira pela qual a escola se organizava e desenvolvia suas ações pedagógicas durante o referido período; no terceiro capítulo, nos aproximamos do pensamento educacional de Lúcia Marques Pinheiro, para, por fim, evidenciar o Serviço de Orientação Psicopedagógico, por intermédio de sua coordenadora Therezinha, e situá-lo dentro da perspectiva de formação de professores que vem sendo discutida no Brasil nos últimos vinte anos. / [en] This dissertation has as objective to understand the Pedagogic and Psycho- Pedagogic Orientations that made possible the construction of the educative experience of the Guatemala School, experimental institution tied with the Departmento de Aperfeiçoamento do Magistério (DAM) of the Centro Brasileiro de Pesquisas Educacionais (CBPE/INEP), with headquarters in Rio de Janeiro, when it was Federal District. For being a research in the field of the History of the Education, elaborated under the perspective of Cultural History, we appeal to interviews and to the documentary analysis of Lúcia Marques Pinheiro and Therezinha Lins e Silva personal files, Pedagogical Supervisor of the INEP and Coordinator of the Serviço de Orientação Psicopedagógico (SOPP), respectively. Starting of a permanent interlocution with Marcus Vinicius Cunha, it was possible to learn the types of orientations that the professors of this school had received during the years from 1955 to 1968, period understood as the end of the movement of the New School in Brazil. If by one side, the pedagogical orientation guided the practical teaching and, in this way, it limited the possibilities of manifestation of the subjectivity of these that taught, of the other side the psychopedagogic orientation provided the creation of a favorable and propitious climate next to the professors so that the pedagogical proposals could be placed in practical. In this way, in the first chapter, we search to characterize the educational and political context that marks the period; in the following one, we describe the way for which this school was organized by itself and developed its pedagogical action during the period; in the third chapter, we approach to the educational thought of Lúcia Marques Pine, for, finally, to evidence the Service of Psycho-Pedagogic Orientation, through its coordinator Therezinha.

Este estudo objetivou investigar a influência da estrutura de propriedade sobre a informatividade dos lucros contábeis divulgados pelas empresas negociadas no mercado acionário brasileiro. Foi considerada como informatividade a intensidade da relação entre o lucro contábil e o retorno das ações, mensurada pelo coeficiente angular da reta estimada entre essas duas variáveis. A pesquisa teve como foco a influência específica de duas características da estrutura de propriedade: a concentração dos votos e a divergência entre os direitos das ações (voto versus fluxo de caixa) sobre a informatividade dos lucros contábeis. Tendo em vista a realidade do mercado brasileiro, a pesquisa considerou a adoção das práticas diferenciadas de governança corporativa como mecanismo redutor da influência da concentração de votos e da divergência entre direitos sobre a informatividade dos lucros contábeis. Adicionalmente, a pesquisa abordou a diferença da informatividade entre o lucro contábil e o dividendo, provocada pela divergência entre os direitos das ações (voto versus fluxo de caixa). A importância deste estudo, no ambiente brasileiro, encontra-se no fato de o mercado brasileiro, diferente dos mercados desenvolvidos, ser caracterizado por uma forte concentração dos votos nas mãos de poucos acionistas e pelo desvio entre os direitos de voto e fluxo de caixa por intermédio da emissão de ações preferenciais. A relação entre a informatividade dos lucros contábeis e a concentração de votos e a divergência entre direitos foi fundamentada pelo efeito entrincheiramento e pelo efeito alinhamento. A metodologia adotada foi baseada na utilizada nos estudos de Fan e Wong (2002) e Francis et. al. (2005) desenvolvidos, respectivamente, no mercado do sudoeste asiático e no mercado americano. Utilizou-se a definição do maior acionista controlador. A amostra utilizada na pesquisa foi formada pelas empresas não-financeiras listadas na Bolsa de Valores de São Paulo (Bovespa) durante o período de 2000 a 2006. Como a amostra de trabalho selecionada apresenta características de um painel desbalanceado, os modelos adotados foram estimados em três abordagens: Pooled, Efeitos Fixos e Efeitos Aleatórios. Os resultados encontrados indicam que a concentração dos votos exerce uma influência negativa sobre a informatividade dos lucros contábeis. Quanto à divergência entre direitos, a influência negativa foi confirmada parcialmente. Diferente do resultado esperado, a divergência entre direitos exerce influência positiva sobre a informatividade dos lucros contábeis. Quando a divergência entre direitos passa a ser observada em conjunto com o excesso de votos, verifica-se a sua influência negativa sobre a informatividade dos lucros contábeis. Essa evidência sugere que a divergência entre os direitos passa a influenciar negativamente a informatividade somente quando o maior acionista controlador possui a maioria (>50%) dos votos. O maior impacto negativo sobre a informatividade dos lucros contábeis foi observado quando ocorre a combinação da concentração de votos com a divergência entre direitos. Sobre a adoção de práticas diferenciadas de governança corporativa, não foi confirmado o seu papel como elemento redutor da influência da concentração de votos e da divergência entre direitos sobre a informatividade dos lucros contábeis. Em relação à comparação da informatividade dos lucros contábeis e dos dividendos, foram levantadas as seguintes evidências: (i) quando condicionadas a divergência entre direitos, foi observada a diferença estatística entre as informatividades dos lucros contábeis e a dos dividendos, e (ii) ao contrário do esperado, sob a influência da divergência entre direitos, a informatividade dos lucros contábeis aumenta e à dos dividendos diminui. / This present study consists on examining the effects of the ownership structure over the informativeness of the accounting earnings which are disclosed by traded companies in the Brazilian stock market. Informativeness is about measuring how intense is the relation between accounting earnings and stock prices considering the angular coefficient of the probable straight line between these two variables. The research focus on the specific influence of two ownership structure´s characteristics: the concentration of the vote rights and the divergence between the rights of the shares (vote versus cash flow) on the accounting earnings informativeness. Analyzing the Brazilian market, this study considerate the use of the different practices of corporative governance to reduce the influence of the concentration of vote rights and the divergence between rights on the accounting earnings informativeness. It also approaches the difference of the informativeness between accounting earnings and dividends, caused by the divergence between the rights of the shares (vote versus cash flow). The relevance of this study, in the Brazilian reality, meets in the fact that the Brazilian market, different from other developed markets, is the typical concentration of the vote rights at the hands of a few shareholders, plus the deviation between the rights to vote and cash flow through emission of preferential shares. The relation between the informativeness of the accounting earnings and the concentration of votes and the divergence among rights was based on the entrenchment and the alignment effects. Its methodology was based on the studies made by Fan and Wong (2002) and Francis et. al. (2005) in Asian southwestern market and the American market. The definition of the largest ultimate owner was used. The sample used in the research was formed by the no-financial companies listed in the São Paulo Stock Exchange (Bovespa) during the years of 2000 the 2006. There were three approaches: pooled, fixed effect and random effect. The results show that the concentration of the vote rights has a negative influence on the accounting earnings informativeness. Concerning the divergence between rights, the negative influence was confirmed partially. However, as an unexpected result, the divergence between rights has positive influence on the accounting earnings informativeness. When the divergence between rights is taken into consideration within the excess of votes, it becomes clear its negative influence over the accounting earnings informativeness. This evidence suggests that the divergence among rights can influence negatively the informativeness only when the largest ultimate owner has the majority (>50%) of the vote rights. The most negative impact on the accounting earnings informativeness was observed when the combination of the concentration of votes with the divergence between rights occurs. Concerning the different practices of corporative governance as the reducing element of the influence of the concentration of votes and the divergence between rights over the accounting earnings informativeness, it was not confirmed. At last, comparing informativeness of accounting earnings and dividends, evidences emerged: (i) when the divergence for rights were conditioned, there was a difference in statistics between informativeness of accounting earnings and dividends, and (ii) whereas, unexpectedally, under the influence of the divergence within rights, accounting earnings informativeness increased and informativeness dividends decreased.

Os efeitos da demonstração na aprendizagem motora tem sido objeto de estudo de longa data. Porém, o efeito da demonstração autocontrolada e das formas de utilização da possibilidade de autocontrolar esse fator ainda são pouco explorados na literatura. Dessa forma, quatro questões de estudo foram formuladas: 1) qual o efeito da demonstração autocontrolada na aprendizagem de uma habilidade motora seriada? 2) há diferença na aprendizagem entre aqueles que gastam mais tempo para tomar a decisão sobre pedir a demonstração e aqueles que gastam menos tempo? 3) há diferença na aprendizagem entre aqueles que processam a informação observada de forma mais rápida e aqueles que processam de forma mais lenta? 4) há diferença na aprendizagem entre aqueles que mais demonstrações pedem e aqueles que menos pedem? Para investigar essas questões, quatros estudos foram realizados com a participação de 120 indivíduos voluntários, universitários, de ambos os sexos. A tarefa de aprendizagem foi uma habilidade motora seriada que consistiu em movimentar com o mouse um quadrado vermelho, exibido na tela de um computador, de forma a tocar sequencialmente todos os alvos retangulares apresentados na tela, em um determinado padrão temporal. O Estudo 1 envolveu a formação de dois grupos: Auto (demonstração autocontrolada, n=60) e Yoked (demonstração espelhada a partir do grupo Auto, n=60). O experimento consistiu de quatro etapas: fase de aquisição (100 tentativas de prática); teste de retenção 24 horas (20 tentativas); teste de transferência 50 (20 tentativas); teste de transferência 100 (20 tentativas). Os Estudos 2 e 3 foram realizados com a participação de 40 e o Estudo 4 com a participação de 38 indivíduos, todos do grupo Auto do Estudo 1. Para cada um desses estudos foram selecionados participantes que obtiveram valores extremos para as variáveis: Tempo para Tomada de Decisão (TTD), Tempo de Processamento da Informação Observada (TPIO) e quantidade de Demonstrações (DEM). Para a verificação da aprendizagem, o Erro Relativo (ER), Erro Total (ET), Erro Absoluto (EA) e Erro na Sequência (ES) foram medidos. Também foram adotadas medidas complementares para a verificação de como os participantes se envolveram no processo. Com relação ao Estudo 1, os principais resultados foram: a) a superioridade do grupo Auto sobre o Yoked, especialmente, quanto ao ER e o EA; b) o grupo Auto processou por mais tempo a informação observada que o grupo Yoked; c) as demonstrações foram requeridas no início da prática, momento em que o desempenho era significativamente baixo, em ambos os grupos; d) o grupo Yoked se mostrou satisfeito com o momento em que recebeu as demonstrações. Os resultados dos Estudos 2, 3 e 4 mostraram, em síntese, que: a) o gasto de mais tempo tomando a decisão sobre pedir a demonstração promoveu melhor aprendizagem que a condição de gasto menor; b) ambos os grupos, formados a partir da variável Tempo de Processamento da Informação Observada (TPIO), aprenderam de forma semelhante; c) o grupo + DEM teve aprendizagem superior ao DEM em todas as medidas de desempenho. A partir desses resultados concluiu-se que a demonstração autocontrolada produziu melhor aprendizagem que a condição externamente controlada, sendo o Tempo para a Tomada de Decisão e a Quantidade de Demonstração, fatores determinantes dessa superioridade / The effects of demonstration on motor learning have been studied for a long time. However, the effect of self-controlled demonstration and the strategies of the learner in this process have been little explored in the literature. Thus, four questions for study were posed: 1) what is the effect of self-controlled demonstration on learning of a serial motor skill? 2) are there learning differences between the learners that spend more time and the learners that spend less time in making a decision about asking for a demonstration? 3) are there differences in the motor learning process between learners that spend more time and the learners that spend less time in processing the demonstration information? 4) are there differences in the motor learning process between learners that ask for more and the learners that ask for less demonstration in the acquisition phase? To investigate these questions four studies were designed involving 120 volunteer university students of both sexes. A timing task was used whose objective was to manipulate the mouse to move a red square on the computer screen so as to touch, in a sequential manner, all the rectangular targets shown on the screen in a determined temporal pattern. Study One was designed with two groups: Self (self-controlled visual model, n=60) and Yoked (control, n=60). The experiment consisted with four stages: acquisition phase (100 practice trials); retention test, 24 hours after acquisition phase (20 trials); transfer test 50 (20 trials); and transfer test 100 (20 trials). The other three studies were performed with participants from the Self group of Study One with 40 subjects being selected for Study Two and Study Three and 38 subjects for Study Four. Subjects were selected according to those that achieved extreme values for the variables of decision-making time, time for processing the demonstration information, and number of demonstrations. To assess motor learning, relative error, total error, absolute error and sequence error were measured. Complementary measures were also adopted to verify how the participants were involved in the process. In relation to Study One, results indicated: a) superiority of the Self group over the Yoked group, especially in relation to relative error and absolute error; b) the Self group processed the demonstration information for a longer time than the Yoked group; c) demonstration was requested at the beginning of practice, a time at which performance was poor for both groups; and d) the Yoked group was satisfied at the time the demonstration was presented. In short, the results of the Studies Two, Three and Four showed that: a) the more time spent in decision making on requesting demonstration, the better the learning of a sequential motor skill; b) the motor learning of both groups formed through time for demonstration information processing was similar; c) the group that asked for more demonstration learned better than the group that asked for less. Based on these results, it may be concluded that the self-controlled demonstration condition promoted better motor learning than the externally-controlled condition, with decision making time and the amount of demonstration being decisive factors in this respect

A ciência (episteme) é entendida por Aristóteles como um conhecimento demonstrativo, isto é, um tipo de saber que pode ser expressado por um discurso (logos) dedutivo fundado em premissas necessárias. No entanto, a demonstrabilidade que caracteriza a ciência não se atribui a seus princípios. Segundo Aristóteles, seria impossível demonstrar absolutamente tudo, pois assim se cairia em uma demonstração infinita e, portanto, tampouco haveria demonstração. Os primeiros princípios das ciências são apreendidos pela inteligência (noûs), a partir de resultados alcançados por indução (epagogé), que é a passagem de particulares a universais. Começando por uma análise dos aspectos formais da ciência, esta tese investiga os diversos sentidos e traços dos processos indutivos, procurando mostrar como eles e outros instrumentos do pensamento podem propiciar conhecimentos seguros que garantam a necessidade do conhecimento científico e de suas demonstrações. / Science (episteme) is understood by Aristotle as a demonstrative knowledge, i.e. a kind of knowledge that can be expressed by a deductive discourse (logos) based on necessary premisses. However, the demonstrability that characterizes science does not apply to its principles. According to Aristotle, it would be impossible to demonstrate everything, since thus we would fall into an infinite demonstration and, therefore, there would be no demonstration at all. First principles of science are grasped by intelligence (noûs), based upon the results achieved by induction (epagogé), which is the proceeding from particulars up to universals. Starting with an analysis of the formal aspects of science, this thesis investigates the various senses and aspects of inductive processes, trying to show how they and other instruments of thought can provide a safe knowledge that ensures the necessity of scientific knowledge and its demonstrations.