5 Examples of How Artificial Intelligence and Machine Learning Are Applied in Neuromarketing

Brain data is an intriguing area to apply artificial intelligence (AI). By measuring brain activity, we can influence consumer behaviour, which is also known as neuromarketing. Consumer psychology is similar, but it focuses more on applying psychological persuasion techniques. The integration of AI and consumer psychology could accelerate the field of neuromarketing in the coming years.  

The development of artificial intelligence within neuroscience has long been stagnant due to the lack of data and computing power. It is very time and labour-intensive to collect a lot of data. Therefore, it was difficult for a long time to create a good predictive model. With an ever-growing network of open science and standardisation of protocols within neuroscience, the use of AI is becoming more common.

Data collection usually occurs in two ways: respondents answer a question after each stimulus, which is later classified by AI, or they are divided into different conditions and classified accordingly.

The applications are endless: for example, in usability, A/B testing, retail research, or detection of cognitive status. There are also countless examples where machine learning is used on brain data for emotion recognition, predicting product preference, or willingness to pay. We will explain a few of these examples.

1. Machine learning for emotion recognition

Machine learning is regularly applied to neurophysiological data for recognising and classifying emotions. This data is collected, for example, while a person looks at a product or image. They then indicate which emotion this evokes, and that emotion is labelled. The model is trained based on that label. Besides looking at products or images, emotions can also be elicited by watching/listening to videos or music. Emotional reactions are increasingly successfully recognised by AI: emotion recognition can - depending on the model, data, and number of classification tasks - predict between 55%¹ and 98%² accurately. The DEAP³ dataset is often used for this, where 32 respondents watch 40 music videos and rate their emotion based on arousal and valence. 

2. Machine learning and product preferences

Research is also conducted into consumer product preferences: After seeing a product, the consumer is asked to indicate whether they like it or not. Artificial intelligence based on this rating is increasingly successful: an unbiased worldview could estimate consumer choice with a 50% chance, machine learning models predict this up to 80%⁴ accurately. 

It is interesting to see the difference between deep learning and machine learning here: an artificial neural network (deep learning) predicted more than 10 percentage points better than the random forest algorithm (machine learning).

3. Machine learning and willingness to pay

Research is also conducted in this way into the optimal price for a product. During our pricing research, for example, the product is first presented to the respondent. Then the brain's reaction to seeing the price is measured, and the respondent is asked if they would buy the product at this price. 

By comparing brain activity during the willingness or unwillingness to buy a product at a certain price and integrating it with the market values of these products, it is possible to establish a predictive model for the ideal price of your product. 

4. Machine learning and attention

AI is also valuable for eye tracking data. We see many applications where a machine learning model is trained based on the eye movements of hundreds of respondents on thousands of photos. This makes it possible to predict eye movements: a heatmap can be created that clearly indicates which elements of the image are likely to attract a lot of attention. 

This is an incredibly powerful tool that, for example, in combination with object recognition (where AI is used to recognise everyday objects, like your FaceID) adds extra value. In this way, we can find out which objects attract attention (such as animals and babies) and which often do not. 

5. Machine learning for detecting cognitive workload

Detecting cognitive workload is an important element within cognitive neuroscience. Originally, this research was aimed at safety, for example, by measuring the mental workload of pilots to determine their optimum: if the work is too easy, mistakes are made because attention has lapsed due to boredom. If the work is too difficult, mistakes are made due to cognitive overload. Berka and colleagues⁵ have also tackled this issue to classify workload. 

From this point in science, classification models have been built for workload, engagement, and distraction, based on machine learning. These metrics are also very important in neuromarketing: when an advertisement is too boring, the message is not remembered, the same applies when it is too difficult. A consumer who is not engaged or very distracted is not valuable for the marketing message. By using the classifications of these cognitive states, we can also identify the highs and lows for your advertising message!

In short; What can AI do for neuromarketing?

• Distinguish between different emotions
• Personalised measurement of your user's internal states
• Predict consumer purchase intention
• Automatically create eye tracking patterns 
• Object recognition in daily life
• And many more.. 

¹Xu, H., & Plataniotis, K. N. (2016, July). EEG-based affect states classification using deep belief networks. In 2016 Digital Media Industry & Academic Forum (DMIAF) (pp. 148-153). IEEE.

² Ramzan, M., & Dawn, S. (2021). Fused CNN-LSTM Deep learning emotion recognition model using Electroencephalography signals. International Journal of Neuroscience, (just-accepted), 1-10.

³ Koelstra, S., Muhl, C., Soleymani, M., Lee, J. S., Yazdani, A., Ebrahimi, T., ... & Patras, I. (2011). Deap: A database for emotion analysis; using physiological signals. IEEE transactions on affective computing, 3(1), 18-31.

⁴ Morillo, L. M. S., Alvarez-Garcia, J. A., Gonzalez-Abril, L., & Ramírez, J. A. O. (2016). Discrete classification technique applied to TV advertisements liking recognition system based on low-cost EEG headsets. Biomedical engineering online, 15(1), 197-218.

⁵ Berka, C., Levendowski, D. J., Lumicao, M. N., Yau, A., Davis, G., Zivkovic, V. T., ... & Craven, P. L. (2007). EEG correlates of task engagement and mental workload in vigilance, learning, and memory tasks. Aviation, space, and environmental medicine, 78(5), B231-B244.