, George Washington University
Pages: pp. 6-8
Abstract—Information technology has not only expanded the scale and scope of global markets, it has also provided the means for probing the meaning of every give-and-take transaction.
The e-mail quickly became a constant nag, a dark angel visitant to remind me of my wrongs. Each time, it asked if I would evaluate my recent automobile purchase at Mr. Tony Pro's Auto Mall. Each time, I replied with a flick of the delete key. I was in no mood to give Mr. Pro any information that might strengthen his hold on me.
Under the best of circumstances, the negotiations over a new car are fraught with inequalities and offer all the advantages to the seller. The sticker price is a fiction. The invoice is no closer to the truth, as a regional office pays incentives on each car sold to better control of flow of products from factory to market. Every bit of information about automotive preferences or driving habits gives a chit of power to the dealer.
Often, the customer has no greater power than bluff and bluster, both of which were factors in my negotiation. However, while most customers buy a car only once every few years—or in my case, 14 years—the seller has the dominant position, having ample experience in offering a ready defense that sports a strong handshake, a firm smile, and a protest that he's doing all he can for you.
So when the honest give-and-take was done, and the new car was parked in front of my house, I was in no mood to give Mr. Pro any more information than was necessary. Although he could protest with all his might that he was merely trying to serve me better, I nevertheless harbored the fear that he was trying to gain the upper hand in a future market transaction.
Information technology has not only expanded the scale and scope of global markets, it has also provided the means for probing the meaning of every give-and-take, for determining if the bargain was valid and if the result was valuable.
Although they're often treated as if they were the same thing, validity and value are very different concepts, Valid activities are those that are done well, that follow basic principles, that demonstrate the truth of the underlying concepts. Valuable activities are those that are important: they expand the field, find application to other disciplines, or generate a flow of money from satisfied customers.
As engineers, we tend to use experts to determine validity and markets to assign value. Experts review new ideas, test their underlying assumptions, and analyze how they were produced. If the results of this effort meet their standards, the experts declare the idea valid. Things can change. The field can evolve. However, as far as the experts can determine, the new idea is a valid addition to the body of knowledge.
In assigning value, markets look beyond validity. An idea can be clever, be completely valid, and represent a substantial intellectual achievement, but still have no value whatsoever. Value need not be measured in terms of the profit a concept can produce in the public market or the numbers of workers its production can employ. An idea can be valuable if it provides tools that expand the field, has applications to other technical subjects, or merely simplifies the task of creating other new ideas.
Yet, markets are influenced by human factors that can temporarily mask value. One party can completely mislead the other into assigning an inflated value to a product. Those who work in the marketplace, such as those who offer automobiles for sale, need to protect their decisions by controlling the flow of information with a paired set of marketing tools: the suggestion system and the recommendation system.
Suggestion systems search for valuable ideas from customers or employees. Recommendation systems present ideas with the intent that customers or employees will find value in them. Both have long histories that could probably be traced, with little difficulty, to transactions on the Silk Road. However, both have been the object of careful analytic study and engineering practice. They show how we systematically try to extract or create ideas.
Suggestion systems are processes that solicit feedback from customers or employees. They're characterized by the familiar, although commonly mocked, suggestion box, usually a container with a slot in its top that accepts slips of paper offering new ideas. With the rise of the modern factory, suggestion systems acquired the trappings of an engineering discipline. They had a theory of operations, a set of best practices, and a professional society, the National Association of Suggestion Systems, which was located in downtown Chicago.
By the middle of the 20th century, the directors of NASS felt that their technology had become crucial to modern industrial management. "Having once and for all demonstrated its undeniable worth," they explained, "the Suggestion System is here to stay, to flourish and thrive and add its full, fair quota to the steady onward march of American Progress through the years."
The standard model for suggestion systems was fairly straightforward. A suggestion box collected ideas from customers or workers. These ideas were sorted by a suggestion clerk, who took them to the appropriate managers for validation. These managers determined if the suggestions were valid and estimated the amount of money that might be earned or saved from each. From this review, the validated suggestions moved to a senior manager or operations committee, who determined the value of each suggestion by looking at the potential income or savings in the context of the entire company. They selected the most valuable suggestions for implementation and rewarded the individuals who had put their ideas in the box.
Despite its confidence in suggestion systems, NASS was forced to admit that suggestion systems often failed to produce any value for the company. These systems were often expensive to operate and often met resistance from managers and engineers.
World War II produced the one brief period when suggestion systems were effective. It was no ordinary time. Many managers were open to the suggestions of outsiders as they were in new leadership roles and had no stake in existing production or operational systems. All felt the urgency of the war. Many, if not most, had relatives or neighbors in combat. By the middle of the conflict, all war production plants were required "to provide machinery whereby each man may submit ideas and suggestions for doing the job better." This requirement came with the promise that these systems would "tap a vast new reservoir of ideas, welding our productive genius into a united effort for victory."
Yet, suggestion systems died a quick public death after the war. NASS declared that "nearly 95 percent of all attempts to operate suggestion systems were unsuccessful." Although the organization continued operations for a decade, it then disappeared.
Suggestion systems soon became objects of ridicule by the public. Cartoons showed suggestion boxes sitting over trash cans, shredders, and toilets. The Computer Society's suggestion box currently sits locked on a lunchroom shelf. No one knows what might be in the box or where the key might be.
The Internet has considerably simplified the process of soliciting and organizing suggestions. Questionnaires can be sent by mail and collated into a central database. Ideas can be sent to managers for required validations. The news of a successful idea can be spread to entire workforces or customer populations. Modern technology has turned the suggestion system into the recommendation system—the software that gathers data on customers and identifies products that each might find valuable.
Recommendation systems, sometimes known as collaborative filtering systems, have proven more valuable than their progenitors. Most large Internet retailers have embraced such systems. It's unusual to purchase items without being told what others have bought, what offerings go well with our selections, or what new product might be of interest to us. One major retailer, Netflix, even sponsored a contest to develop a new recommendation algorithm and, in 2009, rewarded an approach that it identified as a substantial improvement over the prior state of the art.
Recommendation systems have succeeded where suggestions systems did not for a pair of fundamental reasons. First, they address the problems of mass consumption, which are often simpler to grasp and easier to manipulate than the problems of mass production or mass distribution, which were the common focus of suggestion systems. Second, they exploit one of the fundamental strengths of information processing: the ability to record what we actually do, rather than the ideas we propose.
According to a recent survey of the technology, most recommendation systems utilize four types of data. Only the first, rating data, comes from the sort of questionnaires the auto dealership asked me to complete. The remainder is data that describes us—that is, demographic data—or data that describes our actions in the marketplace, such as behavior or transaction data. It's one thing to provide misleading responses on a questionnaire, as it only means that we're lying to others. It's quite something else to engage in misleading market behavior, as such actions mean that we're lying to ourselves.
For the questions from Mr. Tony Pro's business, I provided the most useless of lies. I scrolled through his form and checked the neutral rating of 4 for each query. However, he ultimately turned the tables on me by capturing a little bit of my identity. After completing the questionnaire, his system began to bombard me with notices to have my new car checked by his skilled mechanics. The notices ranged from enticing to threatening, from promising new features for the car to threatening to void the warranty should I drive far more than the engineering specifications recommended.
One day, I received a notice that I would get a free tune-up if I took the car to the shop in the middle of the week. It was a little hook that deftly hid the great truth that there's nothing offered for free in this world that doesn't require payment at some future date. After delivering my car to the dealership, I found a chair in the corner of the showroom, where I began to work on some research on my laptop.
After 20 or 30 minutes, an em-ployee asked me if I would like to use the wireless connection. Grateful for the opportunity, I accepted his offer and then asked him about his job at the dealership. I indicated that I had worked for the auto industry during my college years and was interested in how the work might have changed.
"I'm the data mechanic," my new friend replied.
"Data mechanic?" I asked. "Does that mean you manage the computers and networks in the office?"
"Yes," he responded, "and also the diagnostic and data collection systems in the shop."
A moment passed before I realized what he had told me. My car was sitting in a mechanic's stall, freely volunteering my habits as an owner and driver. It couldn't quite reveal every detail about every trip, but it could download information about the number of miles I had driven, the range of my travels, and the way I operated the car. Perhaps I hit the brakes too hard. Maybe I accelerated too fast from a stopped position. Most likely I dawdled on the freeway, forcing others into a line behind me. All of this information, every little bit, was being relayed to the employees of Mr. Tony Pro.
After this first visit, the e-mails from Mr. Tony Pro changed their tone. Where they once suggested that I might be driving the car too much between visits, they now touted the value of maintaining low-mileage vehicles. He had clearly learned something from my car. He knew that I lived in the city and rarely drove during the week. The information that I had so carefully avoided providing to him in a questionnaire or during an exchange with his employees had been captured from my car's memory and now found a home in Mr. Pro's records. He gained one more chit that day in the economic give-and-take.