Modelling the Creative Processes Behind the Appreciation, Generation, and Criticism of Literature

我的研究目的是模拟故事叙事的欣赏和生成过程，以创造一个代理，至少在一个狭窄的领域内，可以解释故事；创造新的故事；最重要的是，用自然语言解释它自己的解释和创造性的决定。故事是人类认知和交流的核心，处理故事的能力将是人工智能代理的一项重要技能，尤其是与人类互动的人工智能代理。我希望能够帮助理解人工智能如何能够意识到是什么让故事变得有趣和结构良好；能够在创造新内容时借用记忆中的元素；并且可以评估自己的工作。研究的重点更多地放在代理人的自我评价和评论上，而不是高质量文学作品本身的生产。这种关注有三个主要原因：(1)很明显，创造力的核心成分之一是评估人工制品的能力，而不仅仅是生产它们的能力；(2)研究人员越来越认识到需要更多可解释的人工智能模型——一个能够解释其决策的智能体将有助于这一点；(3)一个能够解释自己的智能体需要一定程度的自我感知能力，这种能力可能对未来人工智能和意识领域的工作很重要。我目前正在尝试在一个混合的符号/连接主义系统中工作，因为我相信利用这两种范式的优点并产生一个可解释的模型是很重要的，并且可以作为人类智能的可能解释。

The aim of my research is to simulate the active processes that are involved in the appreciation and generation of story narratives in order to create an agent that can, at least within the confines of a narrow domain, interpret stories; create new stories; and, most importantly, explain its own interpretations and creative decisions in a natural language commentary. Stories are at the heart of human cognition and communication and an ability to deal in stories will be an important skill for an artificially intelligent agent, especially one that interacts with humans.I hope to contribute towards an understanding of how an artificial agent can have an awareness of what makes a story interesting and well-structured; can borrow elements from remembered episodes when creating new ones; and can evaluate its own work. The focus of research is more on the agent's self-evaluation and commentary rather that on the production of high-quality literary artifacts themselves.There are three main reasons for this focus: (1) it is clear that one of the central ingredients for creativity is an ability to evaluate artifacts, not just to produce them; (2) there is a growing recognition among researchers of a need for more explainable artificial intelligence models - an agent that explains its decision making will contribute to this; (3) an agent that explains itself will require some degree of self-perception, an ability which is likely important for future work in artificial general intelligence and consciousness.I am currently trying to work within a hybrid symbolic/connectionist system as I believe it is important to take advantage of the best of both of these paradigms and to produce a model which is interpretable and can serve as a possible explanation of human intelligence.