The confusion of science, engineering and technology

My experience of working with the innovation arms of many equipment vendors and telcos tells me there is a basic confusion around science, engineering and technology. Understanding the difference, and which is limiting progress, is often a key to future success.

“Science is concerned with what is possible while engineering is concerned with choosing, from among the many possible ways, one that meets a number of often poorly stated economic and practical objectives.” — Richard Hamming, Turing Award lecture (1968)

I am finding a recurring pattern in my interactions with consulting clients, be they regulators, equipment vendors, or network operators. The difficulty is one of philosophy: understanding how science, engineering and technology and different things, and why this greatly matters to them.

Consider for a moment the antenna in your mobile phone. The basic ideas around electromagnetism were worked out in the mid 19th to early 20thcenturies. These used mathematics that had been developed in the 17th to 19thcenturies, such as the calculus of integrals, or complex analysis using imaginary numbers. This physics is pure science, and provides a reliable predictive model of an aspect of the universe we live in.

We have understood in principle how radio waves propagate in different media for a long time. In practise, we need to turn that understanding into different mechanisms that perform some useful function in the world.

A clever researcher in the 1990s saw how the reflections of multiple paths could be used to our advantage, and invented the first practical multiple-input and multiple-output (MIMO) antenna. It is quite possible that we might create a model of such a thing before ever successfully building a practical implementation. This model would be applied science to explore the constraints and possibilities of a particular technology domain.

To implement a MIMO antenna in a handset requires many feats: perform the digital calculations fast enough using the available computing machinery, manage power consumption, select appropriate antenna materials, configure them for the radio environment they are most likely to encounter, control size and cost, and so on. This is the process of technological development through engineering. The abstract idea of a MIMO antenna gets turned into a physical device that meets the needs of its maker and user for a specific situation.

Thus far everything is relatively simple. With the digital radio antenna, the mathematics was developed first, then the pure science, which gets turned into applied science that supports the development of specific technologies. The science is a general theory of cause and effect, and technology is a specific configuration of mechanisms designed to solve a particular need.

Digital computing followed a similar pattern. The abstract idea of a computer together with the supporting mathematics was developed in the 1930s, and the first practical implementations followed soon afterwards. With data communications, we have of course been engaged with sending signals over a distance for the whole of human history. However, the information theory of Claude Shannon came in the 1940s and effectively preceded the whole world of transistors and lasers that power the modern datacomms industry.

Now we get to the modern world of telecoms. What data networks do is to copy information between computational processes. This is entirely a business about distributed computing, and we have been doing distributed computing ever since the first paper tape or punched card moved along a corridor between two computer programs.

You can probably already tell what the catch is. The problem is that the technology has run well ahead of the mathematics and science. This is common in the development of knowledge: theory generally often lags practise as we stumble and fumble about.

In particular, the mainstream network engineering community lacks the conceptual tools with which to reason about the performance of complex distributed systems. We build them, we turn them on, and we hope they do roughly what we anticipated. Sometimes we get a nasty surprise when they don’t.

The mathematics and science does exist (see But basic facts about the world of networking are missing from the textbooks. Examples include “quality attenuation is conserved” when scheduling packets, or the existence of two degrees of freedom among load, loss and delay. This means there is a science gap that is constraining the ability of industry players to make progress.

The human and organisational problem I see is perhaps best illustrated with the research and development personnel of the big telecoms equipment vendors. Their staff are individually and collectively rewarded by coming up with technology mechanisms to deliver ever more “bandwidth” at ever lower costs. Their whole existence is to look for the better mousetrap.

In the realm of physics, they can reasonably assume that all the mathematics and pure science they need already exists in an established and mature body of literature. They then falsely make the same assumption about the novel domain of distributed computing, which introduces all kinds of complexity from protocols and algorithmic interactions. This technology is really only about half a century old. Many pioneers are still alive and working hard.

The thought that the R&D teams need to go all the way back to mathematics and pure science is, in my experience, unthinkable to them. What do you mean we need a “complex numbers of stochastics”? That wasn’t on my engineering masters course! And what the hell is this “∆Q” thing you keep on about? I can barely even type it!

You are telling them that if they want to understand how to make a better mousetrap then they need to concern themselves with the physiology of the mouse, the biochemistry of its body, and the basic science of genetics. This is seen as ridiculous, since nobody has had to do that for decades, if ever. Why hunt for something you doubt exists, is needed, or could be readily used?

They are only equipped to deal with applied science, technology and engineering. Their commercial remit is to hunt is for things to patent and trade secrets to develop that will create differentiation. The problem with mathematics and pure science is that you can’t “own” it, and everyone can use it. We no longer inhabit the Bell Labs era of vertically integrated monopolies, so there’s little reward for solving fundamental problems.

Paying too much attention to the “science gap” risks getting fired from your comfy R&D lab job with padded pension. Telling your big boss that you need to go back to work on unpatentable pure maths and science isn’t a career-advancing move. They expect you to be working on more mechanisms and mousetraps!

Indeed, when you tell them they are obsessed with technologies at the expense of the underlying science, you get a blank look. There generally isn’t a need to appreciate the difference, since you see yourself as a technology business, and science is something that happens in academia many steps removed.

So what is the way out? What I have found is that you need to meet people where they are at. Present to them novel technology mechanisms, and explain how they work. Help them see that there are different engineering approaches, and use the new framework as your explanatory guide. Focus on management methods, ignore the mathematics at first.

Only then are they ready and able to engage with bigger concepts like a new science of distributed computing performance. At least, that’s my hope.


For the latest fresh thinking on telecommunications, please sign up for the free Geddes newsletter.