UX as a doctrine. How nineteenth-century Behaviourism still influences UX researchers and designers today.
In this article, we compare current research methods with those facilitated by behaviourist, almost 100 years ago. The similarities are fascinating, but so are the drawbacks that come along with a superficial stimulus-response approach to human behavioural studies. But first, let’s look at the theories behind one of the most influential movements in American psychology.
Behaviourism in a nutshell
The academic field of Psychology was still in its youth at the start of the twentieth century, when Behaviourism more or less evolved out of a deep, reflexive distrust of any behaviour that you could not observe or measure. Some of the most prominent followers of this empiristic scientific doctrine were Edward Thorndike, Ivan Pavlov, John B. Watson and B.F. Skinner. Their behavioural experiments were not motivated by the discovery of what was going on inside the brains or personal lives of their test participants – whether they were a fish, a cat or a human being. All they were interested in observing was what was going on in the outside world before a behaviour occurs and what sort of behaviour these external events or conditions elicit (stimulus-response). The researchers were treating organisms as simple input-output black-boxes according to a ‘just give me something I can see and measure’ rationale. In the mind of behaviourists, people were nothing more than simple mediators between the behaviour and the environment (Skinner, 1993).
This approach became the dominant school in American Psychology by the middle of the last century. Watson suggested, for example, that he is able to turn any child with any background into a doctor, lawyer, artist, thief etc. just by manipulating and controlling the environment the organism lives in (link). Can you make your users do anything you want? I bet you believe so. One of the most cited and ethically controversial experiments of Watson’s was that of Little Albert (Johns Hopkins University, 1920). This experiment demonstrates the empirical evidence of classical conditioning in humans (behaviour is driven by association). B. F. Skinner later invented the Skinner Box which led to several other experiments that exposed the effects of operant conditioning on rats – amongst other animals – in a lab setting (behaviour is driven by consequences).
The key principles of Behaviourism
- Behaviourists assume that all organisms are blank slates to be written on by the physical world around them. They neglect that genes exist, interact and most often get transformed (Epigenetics) by events in our close environment, starting with the prenatal phase. They avoid making cross-references to the ever-transforming biology of our human bodies.
- Behaviourists thought they can produce any behaviour they want, just by exposing their subjects to certain types of stimuli (conditioning). They relied on the idea that human behaviour is shaped by the rewards and punishment of the environment (negative- and positive reinforcement theory).
- Behaviourists were convinced that their approach was universal; that reinforcement or conditioning works in any organism in the same way. You get the same results whether you study it with lab rats, lions in the desert or humans in urban metropoles.
The academic decline in behaviourism
Experts agree that the deepest and most complex reason for Behaviourism’s decline in influence is its commitment to the thesis that behaviour can be explained without reference to non-behavioural mental activities. This, along with its incapability of explaining the human phenomenon of language and memory, build a convincing case against Behaviourism as a comprehensive theory within the academic community of biologists, neuroscientists and psychologists. The psychologist Bill Brewer and the two most loyal Skinner followers Keller and Nancy Breland later demonstrated that conditioning deteriorates after a certain amount of time (Instinctive Drift). Noam Chomsky now argues that conditioning does not exist – it’s just another way of transporting information. Watson later in his academic life got fired from his university and went on to apply his theory successfully in advertising. This probably made it possible to let Behaviourism slip into UX as well.
Even today, behaviourist theories and applications are still popular and continue to be used in the treatment of disorders. They flourish also in the discipline of marketing (Skinnerian Marketing applied by Facebook, Google & Twitter), game development (Gamification) and UX/Usability research. Even I, sometimes, fall into the trap of applying Behaviourism to my user research and design strategies.
Resuscitated applications of behaviourist strategies
Heatmaps – Research by chance
They exist and thrive within the family of those modern semi-automated research tools, solely build to track stimulus-response and treat every user in the same manner, leaving individual human differences outside the discussion and post-test analysis. Why should it matter if someone comes from a socially different background, or has a visual impairment, a short attention span, high cholesterol levels, struggles with dyslexia or has motoric limitations that prevent them from moving the mouse in a precise way? All that counts is how fast determined people can find something on the screen and if there are any conflicting barriers. Since electro-shocks got banned, we now leverage positive reinforcement which kicks in as soon as you found the flashy big red ‘call-to-action’ on the page that rewards you with unveiling the next food trace (or page) inside the operant conditioning chamber. Any behaviourist could not have set up such experiment any better.
Remote, unmoderated usability studies
This is another great research methodology to quickly collect tons of numbers for self-proclaimed statistical relevancy, but not a sufficient way to deeply understand why we do things and how physiological alterations influence human behaviour. Nonetheless, we go for it and select our user research participants more thoughtfully: Male users, aged 45–55 from New York — check. Users who work in offices — check. Users who like meat — check. Since we recruit via Facebook and (still?) have access to some personal, psychographic data we pick only users who are introvert, respond to stress strongly and agree easily. All that’s left is to prepare the operant conditioning chamber (clickable-prototype), the stimulus (we use written words instead of sounds and flashlights) and wait for the response streaming in via automated video and audio recordings. Watson could not have done a better job and would certainly agree with this approach; soon we will let computers fully observe and analyze people’s ‘behaviour’ –cough– I mean responses.
Hold on – have you noticed something? Well, I did. 100 out of our 400 test participants showed varied responses when challenged by the stimulus. A follow-up survey, ethically approved by a recognized institution, unveiled that these 100 participants have a minor or severe mental health handicap. That’s 1 out of every 4. As it turns out, the screener helped recruit quite a few people who previously were diagnosed with post-traumatic stress disorder, depression, alcohol problems, ADHD, OCD, APD, symptoms of anxiety, dyslexia, light schizophrenia, autism and what not. This caused difficulties in text comprehension, decision making, task switching, memorizing tasks or previous selections in a sequence, recognizing visual patterns, resulted in low vigilance etcetera – just to name a few. And surprise – they also came from different social backgrounds. All the things you cannot spot on your radar but heavily influence peoples behaviour when interacting with your system. WOW! That’s quite a lot to take in after you just came back from celebrating your new designs and are still bragging about the fact that you now meet the latest accessibility standards, which are much more inclusive towards people with disabilities. I guess it’s time to ask ourselves: How relevant are the numbers we collected during this research study? And what did we actually learn about our users’ inner workings that would help us predict future behaviour? I think we can now comfortably dump Nielsen’s suggestion to just test with 5 participants unless you want to call yourself a behaviourist.
Usability lab study with eye-tracking, heart rate monitor and a consumer EEG device
We levelled up the game and pulled out the Mercedes under the qualitative UX research methods since we got funding from VCs who know that the market of people who seek ‘mental help’ is getting bi$$er and bi$$er every year. We are a little bit concerned because the app we are about to ship can have a positive or negative impact on people’s mental health, so we just want to get this right. This time, we are more selective with our test participants, acknowledge diversity and let a professional agency recruit the ‘perfect match’. Being a startup, time is not on our site, so we are pleased to get this all done within the next three months. This includes generating a diary study to capture the long term effects of daily app usage with promising numbers flavoured with impactful, self-reported quotes from our best-looking users. Everything is prepared and the Skinner Box is ready to roll; We got a generic stimulus that can work for any human with any background –and – we can record the responses and reward or punish app usage with 140 characters reinforcement notifications. Voila!
On the day of the first usability lab test, everything is going smoothly and the incoming data from the monitoring devices show a calming effect on people’s stress levels (low theta: beta EEG signals in the back of our participant's brain) and consistent mid-low heart rates. Pupil movement looks a bit rapid and uncontrolled with some participants but we can fix this by reducing the number of elements on the screen – so we believe. All in all the early results are Good News we want to immediately share&promote with potential early adopters who cannot wait to download and install another app onto their latest iPhones.
“Hold on,” says my colleague and “let’s pause here for a second: How do you know when a (internal/external) behaviour, observed on one of our test participants has had an influential environmental or social component? And what if the same behaviour has had a biological component on a neuronal, single neuron level?”
I think what she is saying is that the behaviour we see is the behaviour we designed for, but we don’t know what it was that caused this behaviour to occur in the first place. This is quite problematic when you are working on an app that supposedly helps people cure some of their health problems longterm, especially the ones of biological, epigenetic nature. All that we have done with our research so far is fix some app usability problems, collect superficial behavioural data and pretend we conducted scientific research, which we happily shared with our stakeholders and first paying customers. A faux pas like this happened before – welcome to the App Happiness Business.
OK, so what did we just learn:
- Many current UX and usability research methods follow the experimental setups of behaviourists (input-output) and are almost a century old. These assessment techniques rely exclusively on behaviour observation to measure engagement or interaction. They are either too general to grasp how engagement is naturally expressed through human behaviour or too complex to be traced back to an overall engagement state.
- Understanding the ‘inner workings’ of behaviour is difficult and not easy to test for nor easy to ‘see’ (e.g.: what happened a millisecond before an observable behaviour occurred, what happens after?). Hence why, mainstream user researchers are satisfied generating superficial behavioural statistics to improve usage speed, user-friendliness and self-reported system satisfaction. See the latest example from NNgroup.
- It’s hard to test the distribution of cognitive processes across members of a social group, although this would be necessary since they heavily influence research results. Replicating a study result once you change the social setting is hard.
- Certified clinical studies are expensive and most startups cannot afford to dish out 300–500K, just to promote their newly developed app. They opt for the pseudo-science approach for research and investor satisfaction.
- Genes have limited effects on coding behaviour and it’s events in life that nurture the way we engage with the world that surrounds us. These ‘life adaptations’ start at a pre-natal phase and can be ‘passed down’ (are inheritable) up to five generations.
I am curious to hear what you think about all of this? Do you agree or disagree? Add your thoughts and experience to generate a fruitful discussion.
Hi! Did you know I am a freelance UX designer and Researcher? You can hire me for your next research session or product design sprint.You might be interested in reading:
