Part 3: Operant conditioning
Angelo is a retired bus driver, and prides himself on routine. An avid collector, Angelo has religiously bought the same bi-weekly publication his whole adult life, and had done so from the same store, at the same time on release day for as long as he could remember. A year ago, Angelo's regular store was closed for renovation, so he was forced to visit a store on the next block for the month. Upon buying the magazine he explained to Ling, a young worker there, how much the magazine meant to him and that he'd never missed a copy, keeping them safely stowed away in his collection. Two weeks later when he returned, Ling called out to Angelo, asking him to wait a moment. She went into the staff-only area quickly before popping back with a copy of the magazine. "I saved this one for you," she said. "It was the in the best condition of the bunch. I figured I'd keep it safe until you arrived."
From that day forward, Angelo made visiting that store a part of his routine, and Ling would always keep a copy safe for Angelo in the staff area, thanking him for his patronage upon selling him the magazine.
In the above example, Ling used positive reinforcement (adding a desirable outcome to promote recurrence of a behaviour) to encourage Angelo to return regularly to her store. Like discussed in part 1, Ling considered Angelo's intent and facilitated it, pleasing Angelo enough to alter his future behaviour.
Operant conditioning, fathered by B. F. Skinner, is important in wallet design as the vast majority of users will not fully understand the functions, or their implications. It's therefore our duty to design these functions in ways which promote their proper, continued use. For this reason we should design with reinforcement, and not punishment in mind, as the latter may deter users from using the wallet at all. So, wallet design should encourage the voluntary action of the user, should that action align with both the intended design and user intent, by being rewarding. Like Ling, we can use positive reinforcement to achieve this, with the rewarding feedback simply confirmation that the user's intended outcome has been met.
You may be thinking 'confirmation and feedback are rudimentary in design', yet many instances of vague, untimely, or simply nonexistent feedback remain extant in the crypto space. Non-existent feedback is obviously a bad idea as it creates no opportunity to reinforce or punish behaviour; this was a primary tenet of Edward L. Thorndike's pre-Skinnertheory, so let's look at the other two more closely.
Vague feedback would be something like a return of 'null' upon entering a command into a wallet command console, which leaves the user uncertain if they've achieved their goal or not. Imagine taking $5000 USD to the bank to make a deposit into your account. Would you be satisfied if the teller simply took the money, tucked it out of sight, and smiled at you? It's easy to assume the worst, and uncertainty is a powerful stressor, particularly when our hard-earned money is on the line.
Just as banks provide customers with receipts and statements, a wallet should also deliver clear, sufficiently detailed feedback confirming a user's attempted action took place, and that it met their goal. When necessary, a failed action should equally be communicated by the wallet as attempted but failed, with the reason for failure listed and recommended actions to solve the problem.
Untimely feedback would be something too far removed in time from the action for the user to reliably perceive correlation between operant behaviour and consequence—cause and effect. Thorndike posited with his law of effect that this association was integral to learning. Consider the bank deposit example above. How do you expect you'd feel getting a letter after 6 months confirming $5000 USD was successfully deposited into your account? What if you'd tried several different methods of depositing into your account in the intervening months? Could you be sure which particular deposit the letter was in reference to? Although there are many schools of thought on defining causality, and a post hoc correlation-causation logic fallacy, temporal proximity between cause and subsequent effect, known as immediacy, greatly aids inference.
In a wallet, providing the positive feedback directly after the action allows the user to associate their action with the desired result that aligns with their intent. If immediate confirmation is not possible due to confirmations, maturity, or other chain-state reasons (I often see people paranoid while waiting for bitcoin transactions to take place), then some form of progress meter or other clear indicator of the action being underway is essential. This is why electric kettles tend to have an LED that lights up when the kettle is turned on, as bubbling water and rising steam are not immediate consequences of switching the appliance on.
Next we have the consideration of size—cost-benefit analysis on the part of the subject. Size is the judgment made on whether or not the reward outcome justifies the effort necessary to receive it. Imagine meeting a pirate one day and receiving from him a treasure map. He tells you if you can find the X there's another map for you next month, and the month after etc. After navigating through miles of thick jungle to a remote beach, swimming to an island, and digging up a treasure chest, you find a single, small silver piece. How likely are you to take the next map from the pirate?
For a crypto example, sometimes efforts such as chains of 2FA, security e-mails, passwords, capchas etc. before we can finally take a desired action simply make the outcome not worth the trouble. It's important to weigh this up, and identify the point diminishing returns additional security measures aren't needed beyond. The depth at which a particular option or setting is found in a wallet, or the amount of time before a minor action completes are other examples of how size can impact user behaviour negatively against the designer's wishes.
I'll return now to the case of Ling and Angelo. We know that it was only because Angelo's regular store was unavailable that he even stepped into Ling's store in the first place. We also know that Ling recognised that this would bring Angelo back at least one more time, as he would otherwise miss his magazine. This is a case of satiation/deprivation—the measure of which the subject needs the reward—in which Angelo was desperate enough for his magazine to make the extra effort. As this condition was only temporary, Ling knew she had a limited window in which to capture Angelo's continued patronage, so acted while Angelo was in a state of deprivation.
The last factor that influences the power of reinforcement is contingency—the need to remain consistent in outcomes to promote repeat behaviour. Do you think Angelo would still return regularly if Ling only saved him a copy some of the time? If we consider what we know about Angelo, perhaps the size consideration—ie. if the extra distance of the walk was worth it for an uncertainty of payoff—would simply see him returning to the old habit of visiting the closer store.
Remaining consistent in wallet design means a user should reasonably anticipate the outcome of a repeat action to be the same as it was the previous times. This cannot always be the case of course, due to varying factors such as having sufficient resources (funds, computational power), being properly synced, being online etc. These exception should, however, always be intuitive, and as I mentioned earlier in this piece, a failed action should equally be communicated by the wallet as attempted but failed, with the reason for failure listed and recommended actions to solve the problem.
Before wrapping up, I'll briefly mention the other reinforcement type, negative reinforcement, which involves rewarding favourable behaviour by removing an undesirable outcome or stimulus. You may have experienced annoying pop-ups on your computer letting you know updates are available for software. The action which the software developers want you as a user to perform is to update the software. Taking this action will be rewarded with the pop-up alerts stopping. The problem here is that size determination may have the user deem the reward or software as a whole insufficient, and so simply uninstall to end the annoyance. Frequency is a big factor in this.
This was a long one. To summarise, wallet design should encourage the voluntary action of the user, should that action align with both the intended design and user intent, by being rewarding. Positive reinforcement is the best way to achieve this within a wallet. There are four consideration of reinforcement to consider:
- immediacy—are the cause and effect easily linked?
- satiation/deprivation—is there sufficient desire for the outcome?
- contingency—is the outcome consistently linked to the behaviour?
- size—is the outcome sufficient for the effort involved in the behaviour?
Designing with these in mind will greatly increase the chance of teaching a user how to use a wallet, and keep them behaving in ways you intend.