fallacy-of-product-design-in-ai
Georges Adilon, Lycée Sainte Marie-Lyon, 1976). La Verpillière, France. Above: Giuseppe Perugini, Tree house, 1971. Fregene, Italy

Product design can be likened to architecting a set of (friendly and pleasant) rigid corridors. A user walks through these corridors, hopefully reaching their intended goal, while performing actions that can benefit the systems.

We’re entering an age of complexity where not only such user movement can’t be controlled, but the corridors themselves are flexible and adaptive. The entire house (as the containing structure) can disappear and built itself in a generative and adaptive way.

To start explaining what I mean by that I would like to bring in the standing ovation problem (Miller and Page, 2004).

Imagine sitting through one of the most touching and life altering talks of your life.

As the speaker puts the mic down the audience starts clapping, gradually moving into a standing ovation.

The person on your right stands up (she went to high school with the speaker), you hear people cheering behind you on the right (they discussed the topic over dinner before the talk). The speaker bows and more people are standing up.

The entire room is filled with love for the speaker and, you unexpectedly find yourself standing and cheering.

The Standing Ovation problem is often used as a reference to complexity systems, and the edges of what can be modeled.


Now I want to extend it even further.

Imagine sitting in a vantage point and trying to make note of this erratic narrative. You have been tasked with building an AI system to replicate what just happened.

Designing that model is impossible. Even if we could peer into everyone’s brain it would be of little use as we would not know where to look. Let’s say we did collect terabytes of data, and had access to endless computing, we would know what to do computation to.

This situation – on its most basic, and logical level is impossible to model.

And this is an idea worth registering – the Impossibility of modeling.

1 + 1 = 3

Doug Engelbart once said that if you put a group of intellectuals together you get something that is bigger than increased efficiency (Augmenting Human Intellect: A Conceptual Framework, 1962)

The same way that you bringing a group of people into a room, and matching them with a passionate speaker can result in something unplanned, and quite wonderful.

Modern day product design lives and dies by its ability to track, and optimize every user movement.

  • Users drop off at the sign up screen, let’s put an incentive there.
  • We’re not getting enough engagement on this button, let’s change its color.

It’s a reductionist formula that is partly driven by the containment of users, and their actions into business owned silos (more in Gated Products).

But life–as the agents within it – is complex. We are all heterogeneous creatures. We respond to each other, our environment and the context of our thoughts. If we were to design the room (and the people within it) there is absolutely no way to reproduce the standing ovation. Yet we somehow think that we could reduce human ambitions to a sequential set of instructions.


The field of complexity systems was originally born out of economics, and since then has bled into biology, city planning.


Decentralization and Emergence

When we can break down a system into satellite utilities, decentralization gives way to emergence.

That is to say – similarly to our standing audience interactions between the nodes affect the entire system, creating emergent properties.

There is no way of conducting the entire room in an authoritative way, instead the agents (people / users) are left to use the system (the room / your app) as they see fit.

Emergence in software design is an idea I am hoping to look into a lot more.

Nuanced Interfaces: biology not physics

Physics is neat and tidy. Think of Newton’s well–design and all–catching minimal theory of gravity.

Physics deal with the movement of objects in the universe while practicing a level of abstraction. Some details are left out and that is ok.

Biology on the other hand is much messier. It deals with organisms that do the strangest of things. Think of slime mold moving on the floor, its behavior is probably as erratic as the crowed in our Standing Ovation scenario.

We are used to build “physics” inspired, reductionist systems. Systems that are neat and minimal – they make sense, and money (in a way around, VC funded way).

We computed what users might want, how may we produce it and how could we monetize this whole thing.

We’re getting to a place tho, where systems are larger than screens, data is literally everywhere and systems are expected to decipher an array of unstructured inputs (voice, handwriting, images).

In a world where you can speak to a machine, or when it can pick up on your sentiment, a reductionist approach would just not cut it.

Our systems are already complex, but business conventions and user exceptions are holding them in a traditional model.

That is bound to change as we’ll start having sensors in pillows that talks to our shoes – sleep and exercise respectively, and a new wellbeing app that does not hold your data.

This is the reason for the fallacy of product design.

We can’t design a result. We can design an intent.

We can enable our users with the tools, mental model and opportunity to use our system.

We’re entering an age where data is everywhere, as are interfaces.

We will be writing tools in absolutely new ways – funneling human interaction is a fallacy.

Let’s focus on enablement, and clear communication of our system’s purpose and function.