Aware Adoption of Artificial Intelligence and Big Data: a Value Framework for Reusable Knowledge

Artificial Intelligence (AI) provides practical advantages in different applied domains. This is changing the way decision-makers reason about complex systems. Indeed, broader visibility on greater information (re)sources, e.g. Big Data (BD), is now available to intelligent agents. On the other hand, such decisions are not always based on reusable, multi-purpose, and explainable knowledge. Therefore, it is necessary to define new models to describe and manage this new (re)source of uncertainty. This contribution aims to introduce a formal framework to deal with the notion of Value in the AI-BD context, embracing both the multiplicity of Value dimensions and the uncertainty in their visibility as the foundations for a dynamic, relational representation of Value. The framework design is based on abstract and highly scalable definitions to represent Value, even considering the interaction of different agents through comparison, combination, and update of states of knowledge. In such a model, both Big Data and different types of intelligence are considered as resources. The information extracted from data becomes a renewable resource if it can be transformed into knowledge, which is reusable beyond a specific scenario and, dynamically, over time. The focus on reusable knowledge is exploited in the relation between Human and Artificial intelligences, which is characterised by a "non-classical" form of uncertainty related to data observability. Finally, we identify applicative domains for future investigation, in order to address the impact of the dynamic behaviour of Value dimensions on strategies and decision-making, enhancing the adaptability and, hence, the sustainability of AI-BD initiatives over time.

Aware Adoption of AI: from Potential to Reusable Value

Artificial Intelligence (AI) provides practical advantages in different applied domains. This is changing the way decision-makers reason about complex systems. Indeed, broader visibility on greater information (re)sources, e.g. Big Data, is now available to intelligent agents. On the other hand, decisions are not always based on reusable, multi-purpose, and explainable knowledge. Therefore, it is necessary to define new models to describe and manage this new (re)source of uncertainty. This contribution aims to introduce a multidimensional framework to deal with the notion of Value in the AI context. In this model, Big Data represent a distinguished dimension (characteristic) of Value rather than an intrinsic property of Big Data. Great attention is paid to hidden dimensions of value, which may be linked to emerging innovation processes. The requirements to describe the framework are provided, and an associated mathematical structure is presented to deal with comparison, combination, and update of states of knowledge regarding Value. We introduce a notion of consistency of a state of knowledge to investigate the relation between Human and Artificial intelligences; this form of uncertainty is specified in analogy with two scenarios concerning decision-making and non-classical measurements. Finally, we propose future investigations aiming at the inclusion of this form of uncertainty in the assessment of impact, risks, and structural modelling.