Being a spin-off from the computational neuroscience lab at KU Leuven, Mindspeller develops AI-based solutions leveraging years of cutting-edge scientific research lead by our founder, Prof. Dr. Marc Van Hulle.
In his 2016 publication “Single-Trial ERP Component Analysis Using a Spatiotemporal LCMV Beamformer”, Mindspeller’s founder, Prof. Marc Van Hulle and his team at the KU Leuven Neuroscience lab, demonstrated that the association strength for concepts positioned in Mindspeller’s semantic network correlates with N400 EEG measurements.
Prof. Marc Van Hulle
Founder & CSO
First of all, it is important to distinguish traditional neuromarketing approaches that rely on biometric signals (like eye tracking, Galvanic Skin recordings, fMRI scans, EEG brainwave [“rhythms”] analysis) from those that are more scalable. All recordings made on individuals are inherently costly also in terms of analysis, and as a consequence considered for limited cohorts only, potentially leading to wrong conclusions. Why is that? Imagine you want to check how many people own iPhones vs Android smartphones, so you ask a few people on the street - but by chance you just met a lot of people coming out of an Android Fanclub meeting. You might get the notion that no one owns iPhones. Only by measuring a sufficient number of people you will reach conclusions that are more resistant to such odd cases.
Second, it is important to see that many such measurements capture only partially what we are interested in. For example, with eye tracking you can accurately measure where a person is looking at, but you cannot automatically assume their comprehension of what’s in front of their eyes. Anyone who has seen the empty gaze of a bored student during a lecture can attest that. Similarly, simplistic EEG techniques such as brainwave (“rhythm”) analysis are general summations of brain activity, and thus susceptible to drowning the signal of interest. It is enough to see that irrespective of what is happening, closing our eyes will make the alpha wave completely dominate our brainwaves. Moreover, many of the cheap EEG devices do not measure brain activity at all! Instead they measure electrical activity of facial muscles.
The following sketches offer an overview of various traditional neuromarketing techniques based on the recording and analysis of biometric signals.
Galvanic skin response
Heart rate variability
Implicit Association Test
When it comes to EEG, more sophisticated techniques such as ERPs (Event Related Potentials) have been developed. ERPs, as their name suggests, are directly related to certain events. This means that the brain signal can be linked to a sensory stimulus to which the subject is exposed to (the ‘Event’).
Our technique does not suffer from either of the above mentioned problems. The protocol we use is applicable in mass surveys, while its validity and efficacy has been validated with EEG ERP studies conducted by the Computational Neuroscience Lab at KU Leuven and published in peer-reviewed scientific journals. The ERP (Event Related Potential) we used for that - N400 - is directly related to the brain’s semantic processing. It represents a negative deflection (‘N”) peaking around 400 milliseconds after stimulus onset (whence its name, N400), a time frame too short to ‘rationalize’ one’s response and, as such, a relevant proxy of subconscious associations. The N400 is evoked when a concept is not associated with the previous concept or when a concept is not congruent with the context in which it is used.
Various sources in consumer neuromarketing seem to agree that more than 95% of consumer decisions are based on subconscious associations. In fact, the driving authority behind this claim won a Nobel prize for his insights in “system 1” versus “system 2” thinking that lead to this consensus.
Mindspeller develops proprietary semantic networks that leverage the results of the N400 protocol used in a series of EEG and invasive recording studies made in humans.
A semantic network is a knowledge base that represents semantic relations between concepts. Such networks are often used as a form of knowledge representation. Essentially a semantic network is a directed graph consisting of vertices that represent concepts and edges the semantic relations these concepts share (source: https://en.wikipedia.org/wiki/Semantic_network )
To date various EEG studies involving more than 240 EEG test subjects have lead to a validation of our semantic association network as a reliable proxy for mapping collective subconscious associations.
Moreover, our semantic network is based on a free association protocol where the subject is free to enter any association that comes to mind. Contrary to “black box” AI approaches that rely in co-occurrence statistics of words in documents, scientific research lead by Douglas L. Nelson has shown that free association-based techniques are significantly more accurate at predicting recall.
Finally, compared to implicit association tests, in which the subject needs to indicate which one of the shown concepts is best associated with a queried concept, and which are often used and even claimed to be a proxy to subconscious brand positioning, Mindspeller’s neuro-positioning algorithms avoid such forced choices e.g. amongst a set of pre-defined brand attributes.
Instead, once the brand (or related campaign stimulus) is positioned in the Mindspeller network, we can use our neurometric algorithm to compute the semantic distance towards any Intended Association (or benchmark) in the network after the free association data has been collected and the result positioned in the said network.