Les expériences rares, qu’elles soient positives ou négatives, surprennent par leur avènement inopiné et brutal. Toutefois, le plus surprenant reste l’incapacité des organisations à en tirer des leçons appropriées. Les expériences rares challengent les approches traditionnelles de l’apprentissage organisationnel fondées sur la réplication et l’amélioration progressive. De plus, les expériences rares sont souvent complexes : elles sont composées d’une grande variété d’éléments qui interagissent de façon incertaine. De ce fait, elles sont marquées par un niveau élevé d’ambiguïté causale qui peut conduire à un apprentissage superstitieux. Dans ces conditions, la littérature souligne la nécessité de mettre en oeuvre un apprentissage délibéré fondé sur la codification des connaissances. Mais la codification est une arme à double tranchant, qui peut rigidifier l’organisation et la littérature reste assez silencieuse sur comment réaliser une « bonne codification ». Ce travail de thèse s’intéresse alors à la question suivante : comment développer et mettre en oeuvre une stratégie de codification appropriée pour faciliter un apprentissage organisationnel délibéré à partir d’une expérience rare et complexe ? Cette recherche menée au sein de l’Institut de Radioprotection et de Sûreté Nucléaire s’appuie sur étude de cas réaliste critique. Elle vise à étudier deux apprentissages délibérés mis en place au sein de l’institut pour apprendre de deux accidents nucléaires graves : l’accident Fukushima Daiichi et l’accident Three Mile Island. Les résultats obtenus identifient trois mécanismes générateurs clés du processus de codification, leurs modes d’activation et la façon dont ils se combinent. Ces mécanismes s’activent grâce à la fois au contexte environnemental et à l’apparition de structures dédiées à la codification, puis se combinent pour former différentes configurations qui supportent deux cycles distincts d’apprentissage. Ces deux cycles sont essentiels pour apprendre d’une expérience rare et complexe. / Rare experiences, whether they are positive or negative, surprise by their unexpected and brutal occurrence. However, more surprising is organizations’ incapability to draw lessons from such rare experiences. Indeed, these experiences challenge traditional approaches of organizational learning based on replication and incremental improvement. In addition, rare experiences are often complex: they are composed of a large variety of elements that interact in uncertain ways. As a result, rare experiences are characterized by a high level of causal ambiguity that can lead to superstitious learning. In these circumstances, the literature emphasizes the need to implement deliberate learning based on knowledge codification. However, codification is a double-edged sword, which can produce organizational rigidity. Besides, research remains quite silent on how to achieve a "well-performed codification”. This research addresses the following question: how to develop and implement an appropriate codification strategy to facilitate deliberate organizational learning from rare and complex experiences? This research is conducted in the Institute of Radioprotection and Nuclear Safety. It is based on a critical realist case study which aims to study two deliberate learning process implemented within the institute to learn from two serious nuclear accidents: Fukushima Daiichi accident and Three Mile Island accident. Our results identify three key generative mechanisms of the codification process, their activation modes and how they are combined. These mechanisms are activated by both the environmental context and the emergence of dedicated structures to codification. The combination of these mechanisms forms different configurations that support two distinct learning cycles which are essential for learning from a rare and complex experience.

Made available in DSpace on 2015-04-11T14:03:00Z (GMT). No. of bitstreams: 1 Jacilane de Holanda Rabelo.pdf: 4788725 bytes, checksum: 199f5b959e361201af68d3afa0ac710f (MD5) Previous issue date: 2013-02-19 / CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico / The knowledge acquired by the members of a software organization is one of its most important assets. Therefore, the proper management of this knowledge has become a real need in the software development industry. The capture of lessons learned has become a practice developed by some of these organizations in order to manage and transfer knowledge. One of the most common approaches to encourage knowledge transfer is codification. Knowledge codification can facilitate the storage of lessons learned for a future usage. This MA research aims at elaborating a strategy to facilitate the codification of lessons learned in software organizations, consequently promoting this practice. The strategy was denominated PABC-Pattern, and three studies were developed: the first compared the approach initially called ABC-Pattern with another visual approach the Mind Map; the second compared the ABC-Pattern with the codification structure supported by the ACKNOWLEDGE tool. Results from the first study indicated that the participants had a preference in using the ABC-Pattern approach due to its usage and utility. For this preference, it follows some of the reasons identified: a) more codification details; b) a more detailed description of the knowledge s scenario which is more clarifying; c) the possibility of acquiring more information with less effort; and d) the possibility of describing both the problem and solution. The main contribution for the second study were the improvement suggestions for the ABC-Pattern, in which enabled an evolution of the strategy. The second version of the strategy was denominated PABC-Pattern and was experimentally evaluated by a third study, in an observational study, with the help of the Technology Information professionals. The results showed that PABC-Pattern found good acceptance as an approach to codify and share knowledge, and also that industry professionals would recommend its usage where they work / O conhecimento adquirido pelos membros de uma organização de software é um dos seus ativos mais importantes. A gestão adequada deste conhecimento tornou-se uma necessidade real na indústria de desenvolvimento de software. A coleta de lições aprendidas é uma das práticas utilizadas por parte dessas organizações para gerenciar e transferir o conhecimento. Uma das abordagens para incentivar a transferência de conhecimento é a codificação. A codificação do conhecimento pode facilitar o registro das lições aprendidas para sua consulta posterior. O objetivo desta pesquisa de mestrado é elaborar uma estratégia para facilitar a codificação das lições aprendidas em organizações de software, de modo a incentivar esta prática. A estratégia foi nomeada PABC-Pattern. Foram realizados três estudos: o primeiro comparou a abordagem inicialmente chamada ABC-Pattern com outra abordagem visual, o Mapa Mental; o segundo comparou a abordagem ABC-Pattern com a estrutura de codificação apoiada pela ferramenta ACKNOWLEDGE. Os resultados do primeiro estudo indicaram que os participantes tinham uma preferência em utilizar a abordagem ABC-Pattern, devido à sua facilidade de uso e utilidade. Foram identificadas algumas das razões para essa preferência: a) mais detalhes de codificação; b) uma descrição mais detalhada do cenário do conhecimento, que é mais esclarecedora, c) a possibilidade de adquirir mais informações com menos esforço, e d) a possibilidade de descrever tanto o problema e a solução. A principal contribuição do segundo estudo foram as sugestões de melhoria para ABC-Pattern que possibilitaram uma evolução da estratégia. A segunda versão da estratégia foi chamada de PABC-Pattern e foi avaliada experimentalmente através de um terceiro estudo, dessa vez de observação, com profissionais da área de Tecnologia da Informação. Os resultados possibilitaram apontar que PABC-Pattern teve uma boa aceitação como abordagem para codificar e compartilhar o conhecimento e que os profissionais da indústria recomendariam sua utilização na empresa em que trabalham.