DeepDesign. When human-centered design research meets machine learning

Nowadays, user-centered design is widely accepted as a method to tackle challenges and develop products / services / product service systems. The design process at its base is, as defined by the Design Council, a non-linear process that involves the four phases of Discover, Define, Develop and Deliver: the famous double diamond (Design Council, 2005). The core mechanism of the process is to “create choices through a diverging phase [...] and, in convergent phase, do just the opposite: eliminate options and make choices” (Tim Brown, 2009, p. 68), make decisions. Although we could say that “a huge part of user-centered design is following your nose” (IDEO, Field Guide to Human-Centered, Design 2015, p.84), when the discover phase get ten times bigger than normal, it becomes hard to make choices without the risk of arbitrarily getting rid of relevant information, as we will see in the presented case study: CIMULACT. Today’s computational power and availability of an unimaginable amount of data are bringing many other areas of knowledge to the use of machine learning and deep learning, while the design process is not taking advantage of these powerful tools. In this thesis we analyze the application of innovative data analysis tools in the switch between the Discover and Define phases, and how these tools would impact on the whole design process. This analysis lead to the design of a DeepDesign, a B2B product service system aiming to integrate user-centered design research and machine learning to forecast social trends and cultural models around the big themes that today affects our societies.