ACM Interactions

I asked several talented inventors whether there is more to research than invention. It was not a new question for them, but not an easy one either. “Get back to me after the CHI deadline,” said one, immersed in writing papers on his latest inventions.

 “The Edison-Einstein question,” said another.

Where I work, artifact invention and research is the air we breathe. But recall the fish that asked, “What is water?” We may not understand it. What differentiates invention and research?

A stream of invention is the most visible product of computer science. Thanks to Moore’s law and similar legislation, new technologies and new possibilities stream forth as the semiconductor wizards do their work. How does research fit in? We attend research conferences, apply for research funding, and work in research laboratories.

Edison-Einstein may not be germane. Any research conducted by Thomas Edison the inventor could be relevant, but for theoretical physicists, artifacts such as telescopes and accelerators are means, not ends. Some of them grumbled when Ernest Lawrence received a Nobel Prize for inventing the cyclotron. Computer science is different. It explicitly encompasses systems work alongside theory; HCI honors artifact creation, stretching back to Sutherland, Engelbart, and others.

My colleague continued, “There is descriptive research, which may not be associated with an invention. Social media research is descriptive, there is not much invention.”

He was ambivalent about social media research, at one point saying that it was OK to do it as long as one “also does some original research.” Consciously or unconsciously, he did not highly value work that does not include invention.

I continued, “Is invention by itself research, or is something else needed?”

Echoing Edison’s claim that genius is 1% inspiration, 99% perspiration, he replied, “Implementation. Sometimes the invention part only takes 5 minutes, the rest is implementation. And some kind of evaluation and dissemination is needed.”

Technically, you can patent an invention that you can’t implement—for example, an invention could rely on a separate patent for which you do not have a license. However, novel software artifacts are usually implementable. Other types of invention, including patentable process inventions such as freeze-drying and electroplating, are not our focus here. We will consider the roles of evaluation and dissemination after reviewing the concepts invention, science, engineering, design, and research.

Invention

I judged an elementary school Invention Fair in Texas a quarter century ago. My favorite was a teeter-totter. As kids went up and down, it pulled a conveyor belt beneath it on which soda cans were placed and crushed for recycling. Another was a miniature sink attached to and kept under the main sink, with flexible plastic tubes bringing down water so the inventor’s toddler brother could use it to wash his hands, his dishes, “or just play with the water.” A third was simple—scales set to trigger an alarm when the load shifted, a safe place for household guns lying around when children are about.

I loved it. Inventions are memorable. They demo well. And software companies in the reign of Moore’s law need to invent. Necessity being the mother of invention drives a benign cycle. It is rewarded. When you can point to your feature or product, your contribution is tangible. A manager, or a reviewer, does not have to dig to see that something was accomplished. Inventions are very welcome, some of them.

In an earlier post, I observed that creating useful novelty is not easy in the global village. Develop an idea and if one of the seven billion people in the village had the same thought and reached the same conclusion a week earlier, it could already be on YouTube. That said, ever smaller-faster-cheaper electronics will only realize its potential if people invent.

Science

I have also judged science fairs. Young scientists describe a journey, hypotheses and experiments, what turned out as expected and what did not, and their conclusions as to how things work. In contrast, inventors described their inventions and how they might be used. There is overlap, but a different feel: “I discovered this about the world.” versus “I made this for the world.”

Edison: “I never once made a discovery… the results I achieved were those of invention, pure and simple.”

At some level we prize science most highly. The science fair exhibits were good, but science fairs were much less memorable than the invention fair. How might this play out in academia and industry?

Engineering

Engineering as a field is particularly close to invention, not surprising given the potential that is unleashed by semiconductor advances. Hardware and software engineering processes are central to the implementation of our inventions.

Design

We generally consider design to be part of the refinement of an invention, recognizing that artifact design can be innovative.

Research

Research is the broadest term, spanning all disciplines. The meaning of scientific research is relatively well understood, thanks in part to those school science fairs. Research into engineering or design methods is included, but what role does research play in the practice of engineering, design, and invention? What role do engineering, design, or invention play in research? Our initial question was whether invention in and of itself is or isn’t research.

Research into the properties and uses of materials and algorithms is part of engineering. Design can benefit from basic or applied research that identifies situations in which artifacts might be used. Within HCI, the roles of designer, software developer, and user researcher are generally distinct, but the role of user research is to inform design and engineering.

Does evaluating and disseminating an invention make it research?

In some branches of computer science, inventions need not be evaluated to be published in research conferences. Within HCI, the value of a perfunctory “user study” evaluation of systems and design contributions has been debated. For complex systems, realistic short-term evaluations are not always feasible.

Few artifacts described in our research papers or demoed at our conferences are ever disseminated widely—the overwhelming majority of invented HCI artifacts undoubtedly make their final appearance in research conference papers. This is not necessarily a bad thing; it encourages the innovation that we all benefit from cumulatively.

Invention and research

Two extremes: (1) Random mutation and natural selection. Invention is unconstrained, not boxed in by assumptions about what might be useful. Inventions proliferate and the marketplace identifies which are useful. Unexpected successes validate the high failure rate, we hope. (2) Intelligent design. Thorough research into context and risk precedes invention. We get fewer inventions—and it can be a triumph when HCI research eliminates the invention of features or products that would add clutter and no value—but a higher yield of useful outcomes, we hope.

Which is best, or is it somewhere between? 

If the optimal path is somewhere between, we need to confront the fact that research to guide design is more difficult to sell than invention. Reviewers and managers in a field driven by novelty understand “I made this,” but not, “The angry dog you do not hear barking?—I did that.”

Case study: An invention paper and a research paper. Post- PhD, I joined a talented inventive team in a software product development company. We took on several applications or features intended to support groups and solved the technical problems, but the products failed in the marketplace. Few inventions ever prove to be useful, but the educational benefits of failure, although perhaps not zero, are easily exaggerated. I headed back to research to learn why this software species was so challenging. My goal was not to spur innovation. I hoped to figure out what could increase the prospects for success, or identify contexts in which a given invention would more likely succeed. I was moving from (1), unconstrained invention, toward (2), at least some intelligent design.

In 1988, the team I had left released another group support product. Freestyle was lauded by PC Magazine, PC Computing, Computerworld, and Communications Week. Twenty years later, my colleague Bill Buxton marveled that no one had yet replicated Freestyle’s useful, easily understood features. The team published two papers. The invention paper [1] described Freestyle features. The research paper described an unforeseen deployment challenge. A major element in Freestyle’s commercial failure was a mismatch between the nature of significant communication loops, which span organizational units, and prioritizing and budgeting practices, which do not.

A quarter century later, products similar to Freestyle are arriving. The invention paper served its purpose. Despite the evolution of organizational infrastructures over a quarter century, the research paper remains instructive.

The allure of novelty

CHI and related conferences have always focused on new technologies that are or could soon be widely used. We have the opportunity to explore the boundary of invention and research in examining how technology meshes with behavior. That research includes obtaining feedback for iterative design, but can go deeper.

There is pressure to show that our field invents and innovates, perhaps to convince colleagues or funding agencies that we contribute. Artifacts connect. Douglas Engelbart’s obituaries began and often ended with “invented the mouse.” His deeper contributions were more difficult to grasp, such as his emphasis on intelligence augmentation in contrast to artificial intelligence.

Random mutation and natural selection could outperform intelligent design—encourage a million people to invent and we get 1000 useful inventions, benefiting us tremendously, and the 999,000 useless inventions will disappear unlamented. What do you think? It didn’t seem efficient to me in 1988, and it still doesn’t. A useless invention is more likely to be novel—if useful, one of the seven billion other people on the planet would more likely have come up with it already. An invention might be useful in a different context or with minor superficial changes—research could identify the opportunities. I have seen inventions die that had promising untested niches in which they might have thrived.

I have deep appreciation for my inventive colleagues. I have developed a healthy respect for the value of informing their work with research. The forces on scholarship and product development in our field militate against a holistic approach. With limited attention and access to unlimited information, with little time for deep analysis, we look to extract quick takeaways from what we read and see. This is easier with an invention paper or an invention. HCI research conferences may be morphing into invention conferences, with some room for academic papers that focus inwardly on “building theory” and extending the literature. Both signal retreat from HCI’s unique opportunity within computer science to provide direction through a broader perspective.

Endnote:

1. Levine, S.R. & Ehrlich, S.F. (1991). The Freestyle System: A design perspective. In A. Klinger (Ed.), Human-machine interactive systems. Plenum, pp. 3-21.

Thank you to inventive colleagues in jobs past and present, especially Patrick Baudisch, who has long thought deeply about and discussed these issues.

Jonathan Grudin