A Review of Recent Deep Learning Approaches in Human-Centered Machine Learning

Area of Science:

• Artificial Intelligence (AI)
• Machine Learning (ML)
• Human-Computer Interaction (HCI)

Background:

• Deep Learning (DL) has seen a resurgence, driving rapid growth in AI and ML research and applications.
• Increasing human-AI interactions raise concerns about the usability and adoptability of complex DL systems.
• These challenges have led to the emergence of Human-Centered Machine Learning (HCML).

Purpose of the Study:

• To provide an overview and analysis of existing work in Human-Centered Machine Learning (HCML) specifically related to Deep Learning (DL).
• To establish a working definition for HCML through collaboration with domain experts.
• To systematically classify HCML publications and identify research trends, gaps, and future directions.

Main Methods:

• A systematic literature review was conducted to identify relevant publications.
• 162 publications within the HCML field were analyzed and classified.
• Classification criteria included contribution type, application area, and focused human categories.

Main Results:

• A comprehensive analysis of 162 HCML publications related to DL was performed.
• The study classified HCML research based on key aspects, revealing the current landscape.
• Research gaps, conflicting interpretations, and current challenges in HCML were identified.

Conclusions:

• The review provides a foundational definition and classification of Human-Centered Machine Learning (HCML).
• It highlights critical research gaps and challenges, offering a roadmap for future HCML research.
• The findings aim to guide the development of more usable and adoptable AI and ML systems.