Can brainwaves from EEG be used to measure enjoyment of media?

Question

It seems to be very hard to figure out automatically, what content I will enjoy and what content I will not enjoy.

Of course, there are methods that require me to sit in front of a computer and press yes or no when I see content to determine whether I like it or not. But not only is it tedious, it is also not precise. It is not precise, because there are always more emotions involved, it is about the state of emotions and not a single emotion.

Therefore my idea is this:

While I am watching pictures or media of other kind, my brain waves are measured. Having seen a lot of photo, there will be a certain correlation to a photo being seen and a brain reaction. I could use this information to predict which media will evoke certain emotions in me. The only drawback is that this is not a static database that can be used over time but due to my increase of knowledge it needs to be updated. But in general this should give useful information that enables prediction based on machine learning.

Will this approach give me or anyone else performing this experiment useful data?

Answer 1

I think the question whether this experiment will give you useful data can be split into two parts:

1. Does the EEG reflect the emotional state ?
2. Does the stimulus (media,pictures) induce such an emotional state ?

For the first question there is a lot of research going on and this review might be a good starting point to get into the topic. Without knowing much of the literature I would guess though, that you always need some ground-truth data of the emotional state to train your classifier. This would mean that you would need to somehow 'measure' your emotional state during the EEG recording. You could avoid this by separately 'measuring' your emotional state when watching the media but then you would have to assume that it's always the same emotional state which might not be true.

If you don't want to measure the emotional state, you could also try to do unsupervised learning on the EEG signal, maybe saying that you want to find 3 different emotional states and in the best case the algorithm will give you a nice partition of your stimuli into three categories but there is no way that it can know what emotions those categories stand for.

For the second question I guess it would make sense start with a set of stimuli for which you are sure that they induce strong emotions.

All of this is speaking of general emotional states. Reducing the problem to only distinguishing enjoyment vs. no enjoyment would make it a bit easier of course but the aspects to consider remain the same.