Exploring exploratory programming

Abstract: The real-world impacts of social biases in artificial intelligence technologies has come increasingly to the fore in the last several years. Basic comprehensions and translations for how biases are represented in data is seen as a key step forward in mitigating harms in AI products and services. This paper examines the core issues of mental models with users and developers working with AI models, metrics, and interpretability in AI. With the assumption that users of tools such as IBM’s AI Fairness 360 and Google’s What-if Tool work within the environment of computational notebooks, such as those developed by Project Jupyter or Google Colab, this paper looks at the use of notebooks for visualization, collaboration, and narrative. In examining the design implications for these tools and environments, new directions are proposed for the development of more critical interactive tools to empower data science and aI teams to build more equitable AI models in the future.

TL;DR: In this paper, the authors consider the relation between the exploration of new possibilities and the exploitation of old certainties in organizational learning and examine some complications in allocating resources between the two, particularly those introduced by the distribution of costs and benefits across time and space.