Will Code Remain a Relevant User Interface for End-User Programming with Generative AI Models?

Abstract

The research field of end-user programming has largely been concerned with helping non-experts learn to code sufficiently well in order to achieve their tasks. Generative AI stands to obviate this entirely by allowing users to generate code from naturalistic language prompts. In this essay, we explore the extent to which "traditional" programming languages remain relevant for non-expert end-user programmers in a world with generative AI. We posit the "generative shift hypothesis": that generative AI will create qualitative and quantitative expansions in the traditional scope of enduser programming. We outline some reasons that traditional programming languages may still be relevant and useful for end-user programmers. We speculate whether each of these reasons might be fundamental and enduring, or whether they may disappear with further improvements and innovations in generative AI. Finally, we articulate a set of implications for end-user programming research, including the possibility of needing to revisit many well-established core concepts, such as Ko's learning barriers and Blackwell's attention investment model. CCS Concepts:• Human-centered computing → HCI theory, concepts and models; Natural language interfaces; • Computing methodologies → Natural language processing; Neural networks; • Social and professional topics → User characteristics.

"It's like a rubber duck that talks back": Understanding Generative AI-Assisted Data Analysis Workflows through a Participatory Prompting Study

“…The term "generative AI" is extremely broad and encompasses many types of systems [58]. The term can variously refer to core algorithms (e.g., the transformer architecture), specific instantiated models (e.g., , or fully productized systems consisting of an ensemble of models plus additional components (e.g., ChatGPT).…”

“…To provide clarity around this term, Sarkar [58] defines generative AI as "an end-user tool, applied to programming, whose technical implementation includes a generative model based on deep learning". The term "end-user tool" refers to tools that end-users directly interact with, not the underlying algorithms or models.…”

“…An open question in end-user programming research is: to what extent people will still need to write code directly, if generative AI can do this for them from natural language prompts [58]? As generative models advance, the author argues, they may facilitate a significant expansion in the scope and scale of end-user programming activities.…”

“…Data analysis workflows are notoriously tedious, challenging, error-prone, and have high expertise requirements. Generative AI significantly advances the state of the art in facilitating the authoring and debugging of data analysis scripts, reuse of analysis workflows, comprehension of analysis scripts, learning, and exploration [58]. The potential change in user behaviour has been described as the generative shift [58].…”

“…Generative AI significantly advances the state of the art in facilitating the authoring and debugging of data analysis scripts, reuse of analysis workflows, comprehension of analysis scripts, learning, and exploration [58]. The potential change in user behaviour has been described as the generative shift [58]. The generative shift posits three axes of change: intensification (more sophisticated automation will be applied to existing workflows), extensification (more workflows will be automated), and acceleration (workflows which were previously costly will be applied in more contexts, as they become cheaper due to their automation).…”