Optimal Discrepancy

When explaining the familiar outputs of the academic field of ergonomics/human factors, the introductory phrase "XX suited to human characteristics" is often used. For instance, the plastic bottles with an "ergonomic design" advertised in a commercial a few years ago were shaped to fit the human hand's grip. The panels of the ATMs we use in our daily lives are installed at a height suitable for people based on anthropometric data, and the amount of information displayed is set according to the capacity of human short-term memory.

This concept of "XX suited to human characteristics" is now applied not only to physical objects but also to services. Especially in today's world, where data science is prominent, systems that leverage large amounts of data to recommend "XX suited to *your* characteristics" have become widely used in our daily lives. For example, when you attempt to purchase a product on an e-commerce site, you might be recommended items similar to what you are searching for (content-based filtering) or items purchased by consumers similar to you (collaborative filtering). All of these can be considered services based on the premise that things that are "a good fit for me" will satisfy our interests.

Now, this may sound contradictory, but are the things that capture our interest and give us high satisfaction truly those that are a perfect fit for us?

Human interest and curiosity are a type of what is known in psychology as intrinsic motivation. Humans are driven not only by rewards and punishments (extrinsic motivation) but also by motivation where the action itself is the goal (intrinsic motivation). Many experiments have shown that the latter is superior in aspects such as thinking ability, concentration, intuition, and creativity. Regarding this intrinsic motivation, there is an interesting theory called the optimal level of stimulation theory (Berlyne, 1960, 1965, 1971). It posits that intrinsic motivation has an inverted U-shaped relationship with the level of stimulation (Figure 1), meaning that stronger intrinsic motivation is triggered by a level of stimulation that is neither too strong nor too weak—in other words, an optimal level of stimulation. This is easy to understand from our own experiences. For example, when choosing a movie or a novel, we tend to become interested not in genres far removed from our tastes, nor in genres with which we are already very familiar, but in those that are just the right amount of different from our preferences and knowledge, prompting us to think, "Maybe I'll watch this" or "Maybe I'll read this." Similarly, when facing a vaulting box, one that is too high compared to what we can already clear is discouraging, while one at a level we have already mastered offers no challenge. It is the vaulting box set at a height that is optimally discrepant from our current ability that sparks the motivation to say, "Alright, let's give it a shot."

So, what exactly is this "optimal discrepancy"?

In one of our studies, we proposed and tested a method for information design that sustains user interest by quantitatively clarifying this "optimal discrepancy" (*Reference 1).

Today, people increasingly browse the internet without a clear objective. For providers of products and services, this non-purposive information seeking represents a valuable opportunity to provide information to users and stimulate their purchasing intent. Since non-purposive information seeking is based on the user's free will, it can be described as an action based on intrinsic motivation. This raises the question: how can we create opportunities for users to continue their information seeking without disengaging, thereby exposing them to a wide range of information? To address this issue, we formulated a hypothesis based on the optimal level of stimulation theory. We defined a stimulation level scale where highly familiar information represents a very weak stimulus and completely unfamiliar information represents a very strong stimulus. We hypothesized that if we created a site environment where users could continuously encounter information at an optimal stimulation level—somewhere in the middle—their interest would remain high, and their information seeking would persist.

In our experiment, we first observed people's browsing behavior. We had each participant perform a paired comparison evaluation of their familiarity with 24 different information genres. Using this data, we scaled the strength of each participant's interest in each genre on a range of 0 to 1 (using Scheffe's method) (Figure 2). We found that all participants followed a pattern of continuing their information seeking while moving back and forth between weak and strong stimuli (as exemplified in Figure 3). An analysis of the information-seeking transitions of all 15 participants revealed that, when the most familiar genre was set to 0 and the most unfamiliar to 1, the axis of back-and-forth transitions was probabilistically highest at a stimulation level of 0.3–0.4 (Figure 4). Furthermore, brain function measurements taken during browsing showed that the change in Oxy-Hb concentration in the dorsolateral and medial prefrontal cortex—an area thought to reflect the strength of human interest—increased the most when participants were viewing information in genres with a stimulation level of 0.3–0.4. In addition, we found that the conditions under which participants were highly likely not to disengage from browsing were, within a stimulation level range of 0.15–0.8, when the transition difference between the (n-1)th and nth stimulation levels was an increase of 0 to 0.15 or a decrease of -0.5 to -0.35.

Therefore, we can conclude that if we can determine a user's familiarity with various information genres in advance, we can encourage them to voluntarily stay on a site by designing the information flow. This design would be centered on an axis of genres that are about 30–40% discrepant from their known preferences, encouraging them to move back and forth between slightly more familiar and slightly less familiar genres.

Now, these human characteristics can be applied not only to commercial uses but also to human resource development. In another of our studies, we have shown through multiple experiments that these principles can be applied to goal setting in education and training. We defined a scale where the task level a person can currently handle is 0, and a level they feel is "absolutely impossible" is 100. When a goal was set at a task level about 10% higher than their current ability—in other words, at a difficulty level optimally discrepant from their current self—the individual engaged in the training with high motivation, leading to highly effective results.

There are still many more interesting psychological characteristics of humans, and by applying them through engineering, we should be able to make the world a more interesting place. By broadening our horizons and looking at people from different perspectives, we can see a succession of new challenges for new businesses and new methods. I look forward to meeting students, people from industry, and public administrators to share the fascination of this academic field of ergonomics/human factors and experience the joy of solving real-world problems.

*Reference 1

Nakanishi, Miwa, and Motoya Takahashi. "A Psychophysiological Approach to Non-Purposive Information Seeking: Information Design Based on the Optimal Level of Stimulation Theory." *Journal of the Human Interface Society* 21, no. 3 (2019).