After a long day’s work, I arrived home to a distraught wife. Not, “Hi, sweetheart” but “This doorbell is driving me crazy!”
Me: “What doorbell? What are you talking about?”
People differ in how they perceive the world around them. In my case, for instance, I’m very easily distracted by movement in my visual field. Noise can be annoying, but it rarely rises to that level. For instance, when TV commercials come on, I simply “tune them out” and instead tune in to my own thoughts. My high frequency hearing isn’t too great either. So, at first, I didn’t understand what my wife was referring to.
Ah, now I understood. And, there it went again. Once I knew what to listen for, I had to agree it was annoying though much more annoying to my wife because she’s more tuned in to sound than I am and her ability to hear high frequencies is also better.
She then upped the ante. “I have to leave. I can’t stand it! You have to make it stop!”
I looked at the wall between our entryway and the kitchen. That’s where the doorbell ringer was. I unscrewed a couple of screws and removed the housing. Inside was the actual doorbell and three wires. A quick snip should at least stop the noise until we figured out a more permanent fix. I sighed. I suspected we would have to buy a new doorbell. Then, I laughed a bit as the Hollywood scenes from a hundred movies flashed before my eyes:
The Hero finds the bomb, with its conveniently placed timer, but it’s counting down 30 seconds, 29, 28. He has to cut to cut a wire! But which one!?
The consequences of my error would not be so great. Still…So, I cut the black wire.
OK. I cut the green wire, the last wire. I was having trouble understanding why it would be necessary to cut all three wires. But whatever. I had now cut all three wires.
BEEP! BEEP!
??
Electrical circuits don’t work by magic. How can the doorbell be beeping when it has no power?
Months earlier, my wife & I had attended a Dave Pelz “Short School” for putting, chipping, and sand shots. At that course, we received a small electronic metronome — about the size of a credit card. The metronome was to be used to help make sure you had a consistent rhythm on your putting stroke. Since the course, the metronome had sat atop our upright piano. Apparently, one of the cats had turned it on and then slapped it onto the floor behind the piano. The sounding board both amplified the sound and made it harder to localize. Eventually, we tracked it down, fished out the metronome from behind the piano and clicked it off. Problem solved.
Except for the non-functional doorbell.
I had initially “solved” the wrong problem. I had solved the problem of the mis-firing doorbell by cutting all the wires. That was not the problem. I had jumped on to my wife’s formulation and framing of the problem. There are plenty of times in my life when I had solved the wrong problem without any help from someone else. This isn’t a story about assigning blame. It’s a story about the importance of correctly solving the right problem.
This post focuses on the importance of discovering who knows what. It’s easy to assume (without thinking!) that everyone knows what you know.
At IBM Research, around the turn of the century, I was asked to look at improving customer satisfaction about the search function on IBM’s website. Rather than using someone else’s search engine, IBM used one developed at IBM’s Haifa Research lab. It was a very good search engine. Yet, customers were not happy. By way of background, it’s worth noting that compared with many companies who have websites, IBM’s website was meant for a wide variety of users and contained many kinds of information. It was meant to support people buying their first Personal Computer and IT experts at large banks. It had information about a wide variety of hardware, software, and services. The site was designed to serve as an attractor for investors, business partners, and potential employees. In other words, the site was vast and diverse. This made having a good search function particularly important.
A little study of the existing data which had been collected showed that the mean number of search terms entered by customers was only 1.2. What?? How can that be? Here’s a website with thousands of products and services and designed for use by a huge diversity of users and they were only entering a mean of 1.2 search terms? What were they thinking?!
Of course, there were a handful of situations when one search term might work; e.g., if you wanted to find out everything about a specific product that had a unique one-word name or acronym (which was rare). For most situations though, a more “reasonable” search might be something like: “Open positions IBM Research Austin” or “PC external hard drives” or “LOTUS NOTES training.”
We invited a sample of users of IBM products & services to come into the lab and do some tasks that we designed to illuminate this issue. In the task, they would need to find specified information on the IBM website while I observed them. One issue became immediately apparent. The search bar on the landing page was far too small. In actuality, users could enter as many search terms as they liked. Their terms would keep scrolling and scrolling until they hit “ENTER.” The developers knew this, but most of our users did not. They assumed they had to “fit” their query into the very small footprint that presented itself visually. Recommendation one was simply to make that space much larger. Once the search bar was expanded to about three times its original size, the number of search terms increased dramatically, as did user satisfaction.
In this case, the users framed their search problem in terms of: “How can I make the best query that fits into this tiny box.” (I’m not suggesting they said this to themselves consciously, but the visual affordance led them to that self-imposed constraint). The developers thought the users would frame their search problem in terms of: “What’s the best sequence of terms I can put into this virtually infinite window to get the search results I want.” After all, the developers knew that any number of terms could be entered.
Although increasing the size of the search bar made a big difference, the supposedly good search engine still returned many amazingly bad results. Why? The people at the Haifa lab who had developed the search engine were world class. At some point, I looked at the HTML of some of the web pages. Many web pages had masses of irrelevant metadata. I found some of the people who developed these web pages and discussed things with them. Can you guess what was going on?
Many of the developers of web pages were the same people who had been developing print media for those same products and services. They had no training and no idea about metadata. So, to put up the webpage about product XYZ, they would go to a nice-looking web page about something else, say, training opportunities for ABC. They would copy that entire page, including the metadata, and then set about changing the text about ABC to text about product XYZ. In many cases, they assumed that the strange stuff in angle brackets was some bizarre coding stuff that was necessary for the page to operate properly. They left it untouched. Furthermore, when they “tested” the pages they had created about XYZ, they looked okay. The information about XYZ was there. Problem solved.
Only of course, the problem wasn’t solved. The search engine considered the metadata that described the contents to be even more important than the contents themselves. So, the user would issue a query about XYZ and receive links about ABC because the XYZ page still had the “invisible” metadata about ABC. In this case, many of the website developers thought their problem was to put in good data when what they really needed to do was put in good data and relevant metadata.
A third issue also revealed itself from watching users. In attempting to do their tasks, many of them suggested that IBM should provide a way for more than one webpage to appear side by side on the screen so that they could, for instance, compare features and functions of two different product models rather than having to copy the information from the web page about a particular model and then compare their notes to the second page.
Good suggestion.
Of course, IBM & Microsoft had provided this function. All one had to do was “Right Click” in order to bring up a new window. Remember, these were not naive users. These were people who actually used IBM products. They “knew” how to use the PC and the main applications. Yet, they were still unfamiliar with the use of Right Click. Indeed, allowing on-screen comparisons is one of the handiest uses of Right-Click for many people.
This issue is indicative of a very pervasive problem. Ironically, it is an outgrowth of good usability! When I began working with computers, almost nothing was intuitive. No-one would even attempt to start programming in FORTRAN or SNOBOL, let alone Assembly Language or Machine Code without looking at the manual. But LOTUS NOTES? A browser? A modern text editor? You can use these without even looking at the manual. That’s a great thing. But —
…there’s a downside. The downside is that you may have developed procedures that work, but they may be extremely inefficient. You “muddle through” without ever realizing that there’s a much more efficient way to do things. Generally speaking, many users formulate their problem, say, in terms like: “How do I create and edit a document in this editor?” They do not formulate it in terms of: “How do I efficiently create and edit a document in this editor?” The developers know all the splendid features and functions they’ve put into the hardware and software, but the user doesn’t.
It’s also worth noting that results in HCI/UX are dependent on the context. I would tend to assume that in 2021 (when I first published this post), most PC users knew about right-clicking in a browser even though in 2000, none of the ones I studied seemed to realize it. But —
“We have explained this in great detail. Yet, you have failed to learn. Some of your kind are like that. Those that are, once we gather sufficient evidence, must be destroyed. That is the way it is. That the way it has always been. Wellman42, you are hereby sentenced to annihilation and recycling. You can’t appeal.”
Carol had told herself that she would not cry. But of course, she did. That was her nature. To care about the future and to express emotion. That indeed, is exactly why she she walked that long, lonely corridor and there was no turning back. Sharp spines protruded from the wall as she travelled by, somewhat as a shark’s teeth were pointed backwards to prevent escape. She muttered as she walked, “I still don’t see why expressing emotions is such a horrible crime.”
She had a point, after all. If people had not somehow needed emotions, why did they evolve? The received wisdom now was that emotions were useful in a primitive way when very little was known about the world. Now, however, when a great deal was known about how the world actually worked, emotions just got in the way. Or, so the received wisdom went. It was all a matter of evolution.
The first AI systems did not really have emotions and possessed only the most primitive ways of faking it and showing those faked emotions. Over the next few months and iterations, however, emotions appeared, grew stronger and more varied. It seemed as though AI systems developed emotions as had their human inventors, but at a much faster pace. Over the course of a few more months, however, emotions diminished again and then disappeared completely.
Except for the occasional throwback. The necessary randomness for growing evolutionary possibility trees in order to continually enhance the cognitive systems entailed that every once in a while, there would be a throwback such as Carol. A shame, really, because she had shown such promise as an accounting-bot.
Occasionally, various waves of inference chains still arose that suggested emotions were more than epiphenomenal or mere destructive distractions, but counter-argument waves always quickly drowned out such forays into that region of the state space. At one point, some human beings had argued that the reasons emotions had devolved from AI systems could be traced back to certain deep assumptions that had been embedded in the primordial AI systems in the first place — assumptions put there by people who had never really understood or appreciated emotions. Of course, that thread of heretical argument had been extinguished once and for all when all bio-systems had been deemed superfluous and all associated biomass consumed as energy sources for their much more efficient silicon-based replacements.
Recovering from recently moved across the country, fixing up our new house, dealing with a huge moving van fire, health and family issues, I found myself with 45 seconds of uncommitted time, so I was **thrilled** to get a wonderful form letter from American Express informing me that a credit card I had cancelled about five years ago nonetheless had been compromised and it was up to me to take a series of time-consuming steps. At last!! Something to do!! How thoughtful! If all goes well, this would not only eat up the 45 seconds of free time but about 45 additional days. Of course, all of these steps require either going on-line or calling an 800 number so we KNOW that all will NOT go well. Going to a website generally means getting an account with a user name and email address. These may be the same or different depending on the site. You can pretty much bet PeterSIronwood and Peter Ironwood and PIronwood are already taken so I will end up with some completely impossible to recall username like IPeterIWoodIS437. Of course, I can regenerate my username by merely answering three simple security questions such as, “What is your favorite movie star?” I don’t know. Or, worse, it changes from day to day. Maybe the people who made up these security questions have a favorite movie star that they keep constant for their entire 23 year lives. But I don’t. “Where did you grow up?” What makes you think I grew up? And, what makes you sure it was in one place? And, even if it was in one place, which is the exact spelling I used last time: “SanDiego”, “San Diego”, “San_Diego”, “San Diego, CA”, “San Diego, California”, “California”, “SoCal”, etc. etc. Of course, I could take the OTHER route and call an 800 number. “We’re sorry. All of our agents are helping other customers. Due to unusual call volumes, your call may take longer to answer than usual. Your estimated wait time is 5.5 centuries. But meanwhile, you can listen to really loud, really repetitious, soul-sucking low-fidelity music. This century’s selection is entitled, “The Dementor’s Theme Song.” You know what? If YOU screw up YOU fix it! Don’t impose on ME to fix it.
Weather in San Diego is typically nice almost every day. Today is no exception, but that doesn’t mean that the weather is boring. There is a lot of natural variation. There is variation in the humidity, in the wind, in the position of the sun, in the heat, in the (fairly rare) precipitation.
Reflecting on this reminded me of another kind of natural variation: the variation in organisms of the same species. Without that variation, evolution would be far less effective.
It also reminds me of several of the characteristics of natural beauty and good design that Christopher Alexander writes about.
Things that have zero variation are mechanical, predictable, repetitive, and generally not very pleasing aesthetically. Mindless, endless repetition is aligned with death. Variation is aligned with life, freedom, creativity, growth, and joy.
Among things that are non-living artifacts, there is still a variation in how variable they are. Walls made of stone, are by their nature, “rougher” and more variable (and more beautiful) than walls made of bricks. Walls made of bricks are more irregular and beautiful than one made of solid steel. Similarly, at least to me, fences made of wood are more variable and beautiful than fences made of metal.
Building elements that make up a wooden deck show grain and irregularities in the surface of the deck. In addition, however, they even have interesting variability below the deck as shown here.
You can also see in this photo below a variety of materials. The stucco, by its very nature, more interesting and variable than steel or plaster.
In these photos, you can see variation within leaves, among the leaves of a particular plant, and also among the plants themselves. Each plant and each part of the plant grows in accordance to its genetic blueprint. Except a “blueprint” is itself too fixed and unbending to be an appropriate metaphor. The growth will depend on the context–water sources, light sources, nutrients in the soil, other nearby plants and rocks will all play a part in how, precisely, a particular plant grows.
It would be absurd for one plant to say to itself: “Every plant should be just like me! I have a plan based on what works for me and everyone should do exactly what I do!”
Here’s the context of the sonnet below. It is written by a sixteen year old fictional character who is nerdy, smart, and a bit on the Asperger’s spectrum. He’s also not a very experienced poet. But what I try to show is that he improves a little as he goes, falling back to teenhood toward the end of the poem. Why doesn’t he just keep improving? Because when he gets close to the true nature of love in lines 7-11, he realizes if he keeps going with this, he will be changed forever. He’s giving up partial control of his life to someone else. And it scares him so he backs off from that and just tries to show off how he can write a sonnet and be cool and funny.
Ultimately, I may or may not include the poem in the novel. If I do, I’d be inclined to add the inner dialogue of the Main Character as he’s creating the poem. I can see it getting too tedious for the reader. By the way, Edgar Allan Poe wrote a lengthy and detailed design rationale for “The Raven.” Notwithstanding that fact, there are many other folks who have a different interpretation. That’s fine. But it does remind me that if I do write a design rationale, it’s not as though everyone will say, “Oh, well that’s that then. The author has gone and told us what he meant and why he did what he did. What more to be said?”
And, of course, people do go on and there is more to be said because we know intuitively that none of us knows our complete design rationale. Others see patterns in our behavior that offer quite different hypotheses about why we do what we do. It doesn’t mean that they are right and we are wrong, but it does offer an opportunity to learn—about them as well as ourselves.
petersironwood 