Monitoring Emotions in Playtesting
Valence vs. Arousal
So a friend of mine got a hold of a NeuroSky headset and, I gotta be honest here, I turned into an Actual High School Girl because I just like this sort of thing so much.
During the preliminary “Hey lets check this thing out”-session, I noticed that there was an observable correlation between a certain, specific, emotion and what we were able to see on the output, and that was kind of interesting, because I honestly, truly did not expect to see anything like that at this point.
See, noticing a spike in certain variables is one thing, but noticing that it’s tied to a specific emotional state. Well that’s another thing entirely. Allow me to explain.
Why Monitoring Emotions is Difficult
If you read my post over on The Thinking Zygote about the nervous system, or knew about the nervous system already, you know about the sympathetic and parasympathetic nervous systems. If not, here’s a one-sentence review: The sympathetic nervous system is in control of physical changes in your body that prepare you to move or react to stimuli while the parasympathetic nervous system is responsible for bringing you back down.
The fact that we have physical systems in place that kick in under specific conditions, and that they behave in predictable ways means that doctors, scientists, and people like me have a pretty nice set of physical indicators we can track in order to figure out what’s going on with your brain and nervous system.
Your body will never, for example, drop your heart rate in response to you being startled. If it did, then something is seriously wrong.
This gives us a nice starting hypothesis: Because humans are all built about the same way, then the vast majority of humans will exhibit similar physical responses to certain conditions and/or when they are feeling certain emotions, and therefore we can monitor those reactions in order to get information about what people are feeling at a given point in time.
What’s more, if the statement above is correct, we can actually use this same information to modify gameplay itself. A pulse rate monitor integrated with a controller, for example, could be used to provide game AI with the ability to increase the difficulty of the game itself if it sees you are too relaxed at a given point in time.
The problem is that emotions are a bit more complicated than that.
Allow me to present the following scenario:
You’re walking down the street when you see a mammal in front of you. Your mouth goes dry, your heart begins to beat hard, you start to sweat, and suddenly the world literally looks just a little different.
Obviously something has caused your body to prepare itself for action, but if I were a computer monitoring any of those physical responses I would have no idea if you were terrified by the sudden appearance of a grizzly bear between you and the coffee shop, or if you had just encountered the most perfect and beautiful human being you had ever seen and had fallen in love instantly.
Valence vs. Arousal
Emotions are a subjective thing, and while we can measure their magnitude to a certain degree by monitoring your physical responses it’s it’s difficult to tell the difference between a “good” emotion and a “bad” one.
The difference between a dangerous animals and a potential mate in the scenario above demonstrates the difference between these two elements; where “arousal” is the level/amount of physical response and “valence” is the emotional “direction” of that emotion.
Arousal: The physiological and psychological state of being reactive to stimuli. It results in an observable change in the physical state of the body which causes you to become alert and a ready to move and respond.
Valence: This is the fluffy-cognitive-psychology one, and why I keep putting words like “good” in quotes. Valence is the “intrinsic attractiveness or aversiveness” of an emotion. Which is cognitive-psychologist speak for whether or not people would want to feel something. Most people want to be happy, and most people don’t want to be sad, happiness good, sadness bad, happiness positive, sadness negative.
While there is no simple answer of how to monitor valence at the current point in time, there are ways to gain additional information about about user experiences by supplementing arousal data with additional monitoring methods such as recording facial expressions, or asking users to report their emotional states aloud as they test, but these methods are often intrusive, subjective, biased, and let’s face it, people lie.
So, if you’re looking to track and monitor emotions, you need to understand what people’s bodies do, choose an appropriate method of monitoring the variables in which you are interested, and also consider the strengths and limitations of your testing methods themselves.