Send message to Author

Short abstract:

Domestic data are increasingly being used to automate (parts of) everyday routines through machine learning (ML). This technology-driven discourse risks to overlook the complex dynamic nature of domestic life. We promote a shift towards a design-driven discourse through designerly baseline modeling.

Long abstract:

The integration of machine learning (ML) into the analysis of domestic data to predict domestic routines around, for example, climate control, energy usage, and grocery shopping is becoming increasingly prevalent. ML models leverage this data to inform the design of future products aimed at automating (parts of) these practices to varying degrees of accuracy. However, the use of domestic data to develop ML-enabled systems for the home is not without its challenges.

One significant challenge stems from the dynamic nature of households’ everyday lives. Continuous changes in composition, routines, and preferences introduce variability that is difficult to anticipate. Additionally, domestic data gathered from sensors and other quantitative sources, is de-contextualized and thus loses nuance and interpretability when household members are not actively involved in the process. Moreover, as ML systems accumulate data over time, they tend to perform better for common user types and contexts while struggling with less typical ones. As such, ML-enabled systems pose the risk of disrupting how households’ structure their everyday lives at home.

In response to these challenges, there is a growing need for anticipatory design practices. This presentation will discuss the findings of my designerly exploration of algorithmic prototyping to better understand potential disturbances introduced by ML-enabled technologies into everyday domestic life. Lessons learned from this exploration lead to the proposal of a designerly baseline modeling approach. This approach aims to facilitate collaboration between designers and data scientists by combining contextual insights into the complexities of everyday life with quantitative insights into their patterns.