John watched the sun rise across the fields. If this was a normal August day, he’d be appreciating the beautiful land that he and his wife Jenny, had cultivated for more than 40 years. But he’d barely slept, worrying about how to pay the $12,000 bill that they had just received from the hospital after being admitted with tightness in his chest. It was amazing to realize that less than an hour away from his rusty tractors and creaky fences, there were machines that could look deep into his body and people who could forecast whether he’d live or die. But it was also amazing that he could go so quickly from being a cautious man who lived within his means to anxiously accepting a $1,000 per week payment plan that far outstripped the farm’s usual profits. Modern technology had saved his life, but it was also going to cost him the farm, if he couldn’t figure out how to make those payments each week.
Because you all come from different backgrounds, we’ll be using using this farm narrative as a concrete touchpoint throughout the course. In this first week, our hands-on work will focus on getting everybody set up with the technical tools they will need to access the streaming data. This week’s theory reading will focus on the human element of dashboards, and illuminate the context in which they are used.
Dashboards from a user’s prospective
As with any technological tool, a dashboard can only be as valuable as the problem it’s solving. If it’s not solving a valuable problem, or was designed without any user need at all, it will be just as useless as a Punch The Monkey flash ad. So let’s spend a few minutes in Farmer John’s shoes. He has a very clear goal, boosting his farm’s profits to $1,000 per week. This actually puts him ahead of many users, who have only a general desire to be more informed about their situation. We’ll dive into the process of working with that sort of user in future classes.
Since being told to take it easy, John is (unhappily) spending most of his day in the corner of the basement that contains their only computer. He’s trying to find some farmhands by pecking out emails to his local friends. In his mail bin, he’s got last month’s bank statement, a pile of bills, and the old envelope in which he’s estimated his current balance. This represents an unusual user design situation. Most users are in a steady state, where they are fairly comfortable doing most of their routine, even if they achieved that comfort by doing things the inefficient way. John has been forced into a new environment, but is even worse than most when it comes to exploring new tools or techniques. Even with less resistant users, it’s important to understand how they like to take in information normally, and what their physical, mental and emotional situation will be when using your dashboard.
A User’s Decision Making Process
It’s also important to understand their decision making process. The psychology of decision making is still an active area of research, there are a few key takeaways:
It’s important to zoom out from the final moment of the process, which is probably best termed “choosing” rather than “decision making”, and understand that it’s a series of steps, starting with a recognition that any decision needs to be made at all.
People will use different approaches for different decisions. A simple example is choosing a college versus buying a vending machine candy bar. There’s not even a abbreviated version of the pros and cons lists that many people make – it’s simply “I want that. I have a dollar bill. I’m going to buy it”. Even for the same decision, contextual factors can influence a person’s process, as described in the book Thinking, Fast and Slow
Group decision making is a whole different beast (which we won’t focus on in this class)
A User’s Decision Making Biases
Finally, in terms of understanding any user, it’s critical to understand the mental biases that impact all of our decisions, even those of psychologists who study them. Hundreds of them have been shown in psychological studies*, and I’ll describe a few of particular relevance to our work:
- Confirmation Bias: When looking at a jumpy graph with no clear overall moment, you’d be stunned at how many people can see it trending in the direction they want. A mind that comes in with the goal of proving something, which most do (either consciously or not) will almost always manage to find at least one shred of evidence. In fact, a large study showed that a shocking portion of people could ‘remember’ an entirely fictional event, particularly when it supported their political beliefs.
- Superiority Illusion: A Swedish psychological survey found that a staggering 88% of Americans believed themselves to be safer than the median driver. This effect has been replicated, though often with less extreme outcomes, across a variety of self-assessments. It’s quite likely that confirmation bias comes into play here, as people think back to a few of their best moments and decide that they’ve got enough data to safely declare their excellence.
- Hindsight Bias: These previous two biases, applied to past predictions, result in an impressive ability to misremember how accurately we predicted past events. For example, Farmer John might feel that he can predict how the local minor league team is going to do this year after watching one game, because he’s been following this league for more than 20 seasons, and predicted their championship in ’02. This will do no more than annoy Jenny, but his refusal to take expert advice about which crops will grow best (based on one time he was right, 23 years ago) has already cost them serious money. To give a sense of how pervasive this bias is, a study of 82,000 forecasts showed that policy expert’s predictions are often slightly less accurate than a random guess.
Fundamentally, these data dashboards are decision support tools. Such tools must integrate seamlessly into one of the most complex processes of human cognition. This is the root of the complexity that we will be exploring throughout the rest of the course.