What Is the Interface Theory in “Knowledge and Life”?

I kicked off the New Year by reflecting on my time in Sweden last year.

This self-reflection also got me thinking about my ongoing PhD project on human vision. The idea of seeing fascinates me – we observe, perceive, and even can enhance our vision functionally. Yet, we don’t fully understand how our eyes focus or why we prioritize some things over others.

While at the MISMaP office, I came across an article about Interface Theory in “Knowledge and Life”. This caught my interest, and I decided to explore it further. Unlike past theories, Interface Theory focuses more on the evolutionary aspects of perception – an approach that resonates with me.

Before diving into Interface Theory’s main idea, let’s compare it with the established views. Traditionally, the following principles have been accepted:

What Do the Traditional Principles of Perception Say?

1. Principle of Faithful Depiction: Perception aims to accurately recover or estimate objective properties of the physical world.
2. Principle of Perceptual Categorization: Categorization aims to estimate the objective, statistical structure of the physical world.

These principles are based on Bayesian theory. This guides how probabilistic functions like perceived shadows or texture gradients recreate images that closely resemble reality.

According to Knill and Pouget (2004), the brain represents sensory information probabilistically. Furthermore, it combines this with prior knowledge in a manner consistent with Bayesian inference.

The theory posits that from childhood, we learn to fill in gaps in images. As adults, we automatically complete missing parts using prior knowledge.

Vision theorists widely accept this paradigm. Interface Theory introduces a new principle: the Principle of Satisficing Categories.

This holds that each perceptual category of an organism, shaped by natural selection, serves as a satisficing solution to adaptive problems. According to Simon (1956), organisms with bounded rationality do not optimize exhaustively.

Instead, they settle for solutions that are “good enough” given the constraints of their environment, a strategy he termed satisficing.

Why Is Fitness the Supreme Cost Function?

Additionally: Fitness, not accuracy, is the goal optimized by evolution.

Fitness ties in with game theory principles, like the number of offspring produced. Interface Theory argues that human perception doesn’t primarily aim to reproduce accurate images, but rather to maximize utility. Essentially, “having as many children as possible” matters more than perceiving absolute truth.

This leads to Interface Theory’s core idea. An organism’s perceptions act as a user interface connecting it with the objective world.

In essence, it’s not about reconstructing an image but about processing it into useful and finite information. According to Hoffman, Singh, and Prakash (2015), perception did not evolve to reveal objective reality. Rather, it evolved to guide adaptive behavior, much like a computer interface hides the underlying code while still letting the user act effectively.

Usefulness is relative, creating different interfaces for each species.

Why Do Organisms Rely on Interfaces at All?

Why use interfaces? Processing every bit of information uses up energy and time, so simplification is necessary.

For example, an evolutionary game where players choose a territory from three options with varying food and water illustrates this. An individual who determines if an area is better or worse than average is more efficient than one who precisely measures the resources in each area.

According to Mark, Marion, and Hoffman (2010), evolutionary game simulations show something significant. Organisms tuned to fitness rather than to objective truth can consistently outcompete and drive to extinction those tuned to perceive reality accurately.

The theory fascinates me, though I’m more drawn to the core idea than the subsequent model. While emotions and instincts shape perception, it doesn’t mean humans operate like computers treating encountered objects as mere files.