PyFlowML: A Visual Language Framework to Foster Participation in ML-Based Decision Making

Nowadays, Artificial Intelligence (AI) has become ubiquitous, with machine learning (ML) systems – rooted in algorithms and statistical models – playing a pivotal role in societal advancement. Nevertheless, given their complexity, less-experienced computing practitioners depend on skilled professionals. This paper addresses the challenge of democratizing AI, emphasizing the need for inclusivity in ML-based system design. We introduce the PyFlowML prototype, which extends the PyFlow framework, to investigate how Visual Programming Languages (VPLs) and no-code platforms can enhance user engagement in ML. PyFlowML tailors visual scripting capabilities to meet the specific needs and complexities inherent in ML analysis. This preliminary study is based on a heuristic evaluation of PyFlowML’s usability: we analyzed expert interactions with the tool, exploring its features to design trustworthy ML-based prototypes using Explainable AI (XAI) techniques. We employed the cognitive walk-through method, wherein experts engaged in a series of activities with PyFlowML while sharing their thoughts in real time during the session. While initial findings are promising, they also indicate that to effectively lower the entry barrier to ML-based system design and encourage broader participation, it is crucial to implement strategies that reduce the inherent complexity of ML analysis. This research sets the groundwork for future exploration into how VPL-based tools can transform the design of ML-based systems, aiming for more inclusive and collaborative AI development.