Publications
Schamp, C., Hartmann, J., & Herhausen, D. (2024). Bye-bye Bias: What to Consider When Training Generative AI Models on Subjective Marketing Metrics. NIM Marketing Intelligence Review, 16(1) 42-48. https://doi.org/10.2478/nimmir-2024-0007
2024
Jochen Hartmann,
Dennis Herhausen
Bye-bye Bias: What to Consider When Training Generative AI Models on Subjective Marketing Metrics
Abstract:
Biased training data can distort the outputs of GenAI models and should therefore be a core concern when the models are developed. Ideally, potential biases should be assessed and addressed before the model training and complement the current practice of error-analysis in post-training. This will make the AI model training not only more effective but also more cost-efficient. Being aware of the most relevant biases—sampling bias, measurement bias, social desirability bias and response bias—is therefore essential. Setting up diverse research teams with both technical and market research skills helps to combine these perspectives and allows for the development of successful use cases, especially for subjective training tasks that aim to capture relevant marketing metrics in a specific marketing context.
Authors
- Christina Schamp, Professor of Marketing, Institute of Digital Marketing and Behavioral Insights, Vienna University of Economics and Business
- Jochen Hartmann, Professor of Digital Marketing, GenAI Lab, Technical TUM School of Management,University of Munich
- Dennis Herhausen, Professor of Marketing, School of Business and Economics, Vrije Universiteit Amsterdam
Other articles of the MIR issue “Generative AI - Reshaping the Marketing Landscape”
Here you can find more exciting articles of this issue.
