Assessing the Effectiveness of Textual Recommendations in KoopaML

Abstract

Artificial intelligence (AI) integration, notably in healthcare, has been significant, yet effective implementation in critical areas requires expertise. KoopaML, a previously developed visual platform, aims at bridging this gap, enabling users with limited AI knowledge to build ML pipelines. Its core is a heuristic-based ML task recommender, offering guidance and contextual explanations. The authors compared the use of KoopaML with two non-expert groups: one with the recommender system enabled and the other without. Results showed KoopaML's intuitiveness benefits all but emphasized that textual guidance doesn't substitute for fundamental ML understanding. This underscores the need for educational components in such tools, especially in critical fields like healthcare. The paper suggests future KoopaML enhancements include educational modules, making ML accessible, and ensuring users develop a solid AI foundation. This is crucial for quality outcomes in sectors like healthcare, leveraging AI's potential through enhanced non-expert user capability.