Linking Human And Machine Behavior: A New Approach to Evaluate Training Data Quality for Beneficial Machine Learning

Area of Science:

• Machine Learning Ethics
• Data Quality Assessment
• Human-Computer Interaction

Background:

• Machine learning performance is heavily influenced by data quality.
• Current data quality metrics lack ethical considerations, despite the critical role of training data.
• Ethical dimensions of data quality are essential for responsible AI development.

Purpose of the Study:

• To introduce novel ethical dimensions for assessing data quality in supervised machine learning.
• To propose a new framework for selecting training data based on ethical evaluations of user behavior.
• To shift from a 'big data' approach to a more selective data processing methodology.

Main Methods:

• Analysis of human-computer interaction patterns based on social and psychological factors.
• Development of an ethical assessment framework for behavioral data.
• Proposal of an innovative data filtering regime for machine learning training sets.

Main Results:

• Identified new ethical dimensions of data quality beyond technical metrics.
• Demonstrated the social relevance of varying data qualities in machine learning development.
• Established a conceptual filter regime for ethically sourced training data.

Conclusions:

• Ethical assessment of data is crucial for developing beneficial machine learning applications.
• A selective approach to training data, guided by ethical principles, is necessary.
• This research promotes responsible AI development for both industry and academia.