The Design of Everyday Things

Home > Other > The Design of Everyday Things > Page 12
The Design of Everyday Things Page 12

by Don Norman


  Short-term memory is invaluable in the performance of everyday tasks, in letting us remember words, names, phrases, and parts of tasks: hence its alternative name, working memory. But the material being maintained in STM is quite fragile. Get distracted by some other activity and, poof, the stuff in STM disappears. It is capable of holding a postal code or telephone number from the time you look it up until the time it is used—as long as no distractions occur. Nine- or ten-digit numbers give trouble, and when the number starts to exceed that—don’t bother. Write it down. Or divide the number into several shorter segments, transforming the long number into meaningful chunks.

  Memory experts use special techniques, called mnemonics, to remember amazingly large amounts of material, often after only a single exposure. One method is to transform the digits into meaningful segments (one famous study showed how an athlete thought of digit sequences as running times, and after refining the method over a long period, could learn incredibly long sequences at one glance). One traditional method used to encode long sequences of digits is to first transform each digit into a consonant, then transform the consonant sequence into a memorable phrase. A standard table of conversions of digits to consonants has been around for hundreds of years, cleverly designed to be easy to learn because the consonants can be derived from the shape of the digits. Thus, “1” is translated into “t” (or the similar-sounding “d”), “2” becomes “n,” “3” becomes “m,” “4” is “r,” and “5” becomes “L” (as in the Roman numeral for 50). The full table and the mnemonics for learning the pairings are readily found on the Internet by searching for “number-consonant mnemonic.”

  Using the number-consonant transformation, the string 4194780135092770 translates into the letters rtbrkfstmlspncks, which in turn may become, “A hearty breakfast meal has pancakes.” Most people are not experts at retaining long arbitrary strings of anything, so although it is interesting to observe memory wizards, it would be wrong to design systems that assumed this level of proficiency.

  The capacity of STM is surprisingly difficult to measure, because how much can be retained depends upon the familiarity of the material. Retention, moreover, seems to be of meaningful items, rather than of some simpler measure such as seconds or individual sounds or letters. Retention is affected by both time and the number of items. The number of items is more important than time, with each new item decreasing the likelihood of remembering all of the preceding items. The capacity is items because people can remember roughly the same number of digits and words, and almost the same number of simple three- to five-word phrases. How can this be? I suspect that STM holds something akin to a pointer to an already encoded item in long-term memory, which means the memory capacity is the number of pointers it can keep. This would account for the fact that the length or complexity of the item has little impact—simply the number of items. It doesn’t neatly account for the fact that we make acoustical errors in STM, unless the pointers are held in a kind of acoustical memory. This remains an open topic for scientific exploration.

  The traditional measures of STM capacity range from five to seven, but from a practical point of view, it is best to think of it as holding only three to five items. Does that seem too small a number? Well, when you meet a new person, do you always remember his or her name? When you have to dial a phone number, do you have to look at it several times while entering the digits? Even minor distractions can wipe out the stuff we are trying to hold on to in STM.

  What are the design implications? Don’t count on much being retained in STM. Computer systems often enhance people’s frustration when things go wrong by presenting critical information in a message that then disappears from the display just when the person wishes to make use of the information. So how can people remember the critical information? I am not surprised when people hit, kick, or otherwise attack their computers.

  I have seen nurses write down critical medical information about their patients on their hands because the critical information would disappear if the nurse was distracted for a moment by someone asking a question. The electronic medical records systems automatically log out users when the system does not appear to be in use. Why the automatic logouts? To protect patient privacy. The cause may be well motivated, but the action poses severe challenges to nurses who are continually being interrupted in their work by physicians, co-workers, or patient requests. While they are attending to the interruption, the system logs them out, so they have to start over again. No wonder these nurses wrote down the knowledge, although this then negated much of the value of the computer system in minimizing handwriting errors. But what else were they to do? How else to get at the critical information? They couldn’t remember it all: that’s why they had computers.

  The limits on our short-term memory systems caused by interfering tasks can be mitigated by several techniques. One is through the use of multiple sensory modalities. Visual information does not much interfere with auditory, actions do not interfere much with either auditory or written material. Haptics (touch) is also minimally interfering. To maximize efficiency of working memory it is best to present different information over different modalities: sight, sound, touch (haptics), hearing, spatial location, and gestures. Automobiles should use auditory presentation of driving instructions and haptic vibration of the appropriate side of the driver’s seat or steering wheel to warn when drivers leave their lanes, or when there are other vehicles to the left or right, so as not to interfere with the visual processing of driving information. Driving is primarily visual, so the use of auditory and haptic modalities minimizes interference with the visual task.

  LONG-TERM MEMORY

  Long-term memory (LTM) is memory for the past. As a rule, it takes time for information to get into LTM and time and effort to get it out again. Sleep seems to play an important role in strengthening the memories of each day’s experiences. Note that we do not remember our experiences as an exact recording; rather, as bits and pieces that are reconstructed and interpreted each time we recover the memories, which means they are subject to all the distortions and changes that the human explanatory mechanism imposes upon life. How well we can ever recover experiences and knowledge from LTM is highly dependent upon how the material was interpreted in the first place. What is stored in LTM under one interpretation probably cannot be found later on when sought under some other interpretation. As for how large the memory is, nobody really knows: giga- or tera-items. We don’t even know what kinds of units should be used. Whatever the size, it is so large as not to impose any practical limit.

  The role of sleep in the strengthening of LTM is still not well understood, but there are numerous papers investigating the topic. One possible mechanism is that of rehearsal. It has long been known that rehearsal of material—mentally reviewing it while still active in working memory (STM)—is an important component of the formation of long-term memory traces. “Whatever makes you rehearse during sleep is going to determine what you remember later, and conversely, what you’re going to forget,” said Professor Ken Paller of Northwestern University, one of the authors of a recent study on the topic (Oudiette, Antony, Creery, and Paller, 2013). But although rehearsal in sleep strengthens memories, it might also falsify them: “Memories in our brain are changing all of the time. Sometimes you improve memory storage by rehearsing all the details, so maybe later you remember better—or maybe worse if you’ve embellished too much.”

  Remember how you answered this question from Chapter 2?

  In the house you lived in three houses ago, as you entered the front door, was the doorknob on the left or right?

  For most people, the question requires considerable effort just to recall which house is involved, plus one of the special techniques described in Chapter 2 for putting yourself back at the scene and reconstructing the answer. This is an example of procedural memory, a memory for how we do things, as opposed to declarative memory, the memory for factual information. In both cases, it can take considerable time
and effort to get to the answer. Moreover, the answer is not directly retrieved in a manner analogous to the way we read answers from books or websites. The answer is a reconstruction of the knowledge, so it is subject to biases and distortions. Knowledge in memory is meaningful, and at the time of retrieval, a person might subject it to a different meaningful interpretation than is wholly accurate.

  A major difficulty with LTM is in organization. How do we find the things we are trying to remember? Most people have had the “tip of the tongue” experience when trying to remember a name or word: there is a feeling of knowing, but the knowledge is not consciously available. Sometime later, when engaged in some other, different activity, the name may suddenly pop into the conscious mind. The way by which people retrieve the needed knowledge is still unknown, but probably involves some form of pattern-matching mechanism coupled with a confirmatory process that checks for consistency with the required knowledge. This is why when you search for a name but continually retrieve the wrong name, you know it is wrong. Because this false retrieval impedes the correct retrieval, you have to turn to some other activity to allow the subconscious memory retrieval process to reset itself.

  Because retrieval is a reconstructive process, it can be erroneous. We may reconstruct events the way we would prefer to remember them, rather than the way we experienced them. It is relatively easy to bias people so that they form false memories, “remembering” events in their lives with great clarity, even though they never occurred. This is one reason that eyewitness testimony in courts of law is so problematic: eyewitnesses are notoriously unreliable. A huge number of psychological experiments show how easy it is to implant false memories into people’s minds so convincingly that people refuse to admit that the memory is of an event that never happened.

  Knowledge in the head is actually knowledge in memory: internal knowledge. If we examine how people use their memories and how they retrieve knowledge, we discover a number of categories. Two are important for us now:

  1.Memory for arbitrary things. The items to be retained seem arbitrary, with no meaning and no particular relationship to one another or to things already known.

  2.Memory for meaningful things. The items to be retained form meaningful relationships with themselves or with other things already known.

  MEMORY FOR ARBITRARY AND MEANINGFUL THINGS

  Arbitrary knowledge can be classified as the simple remembering of things that have no underlying meaning or structure. A good example is the memory of the letters of the alphabet and their ordering, the names of people, and foreign vocabulary, where there appears to be no obvious structure to the material. This also applies to the learning of the arbitrary key sequences, commands, gestures, and procedures of much of our modern technology: This is rote learning, the bane of modern existence.

  Some things do require rote learning: the letters of the alphabet, for example, but even here we add structure to the otherwise meaningless list of words, turning the alphabet into a song, using the natural constraints of rhyme and rhythm to create some structure.

  Rote learning creates problems. First, because what is being learned is arbitrary, the learning is difficult: it can take considerable time and effort. Second, when a problem arises, the memorized sequence of actions gives no hint of what has gone wrong, no suggestion of what might be done to fix the problem. Although some things are appropriate to learn by rote, most are not. Alas, it is still the dominant method of instruction in many school systems, and even for much adult training. This is how some people are taught to use computers, or to cook. It is how we have to learn to use some of the new (poorly designed) gadgets of our technology.

  We learn arbitrary associations or sequences by artificially providing structure. Most books and courses on methods for improving memory (mnemonics) use a variety of standard methods for providing structure, even for things that might appear completely arbitrary, such as grocery lists, or matching the names of people to their appearance. As we saw in the discussion of these methods for STM, even strings of digits can be remembered if they can be associated with meaningful structures. People who have not received this training or who have not invented some methods themselves often try to manufacture some artificial structure, but these are often rather unsatisfactory, which is why the learning is so bad.

  Most things in the world have a sensible structure, which tremendously simplifies the memory task. When things make sense, they correspond to knowledge that we already have, so the new material can be understood, interpreted, and integrated with previously acquired material. Now we can use rules and constraints to help understand what things go together. Meaningful structure can organize apparent chaos and arbitrariness.

  Remember the discussion of conceptual models in Chapter 1? Part of the power of a good conceptual model lies in its ability to provide meaning to things. Let’s look at an example to show how a meaningful interpretation transforms an apparently arbitrary task into a natural one. Note that the appropriate interpretation may not at first be obvious; it, too, is knowledge and has to be discovered.

  A Japanese colleague, Professor Yutaka Sayeki of the University of Tokyo, had difficulty remembering how to use the turn signal switch on his motorcycle’s left handlebar. Moving the switch forward signaled a right turn; backward, a left turn. The meaning of the switch was clear and unambiguous, but the direction in which it should be moved was not. Sayeki kept thinking that because the switch was on the left handlebar, pushing it forward should signal a left turn. That is, he was trying to map the action “push the left switch forward” to the intention “turn left,” which was wrong. As a result, he had trouble remembering which switch direction should be used for which turning direction. Most motorcycles have the turn-signal switch mounted differently, rotated 90 degrees, so that moving it left signals a left turn; moving it right, a right turn. This mapping is easy to learn (it is an example of a natural mapping, discussed at the end of this chapter). But the turn switch on Sayeki’s motorcycle moved forward and back, not left and right. How could he learn it?

  Sayeki solved the problem by reinterpreting the action. Consider the way the handlebars of the motorcycle turn. For a left turn, the left handlebar moves backward. For a right turn, the left handlebar moves forward. The required switch movements exactly paralleled the handlebar movements. If the task is conceptualized as signaling the direction of motion of the handlebars rather than the direction of the motorcycle, the switch motion can be seen to mimic the desired motion; finally we have a natural mapping.

  When the motion of the switch seemed arbitrary, it was difficult to remember. Once Professor Sayeki had invented a meaningful relationship, he found it easy to remember the proper switch operation. (Experienced riders will point out that this conceptual model is wrong: to turn a bike, one first steers in the opposite direction of the turn. This is discussed as Example 3 in the next section, “Approximate Models.”)

  The design implications are clear: provide meaningful structures. Perhaps a better way is to make memory unnecessary: put the required information in the world. This is the power of the traditional graphical user interface with its old-fashioned menu structure. When in doubt, one could always examine all the menu items until the desired one was found. Even systems that do not use menus need to provide some structure: appropriate constraints and forcing functions, natural good mapping, and all the tools of feedforward and feedback. The most effective way of helping people remember is to make it unnecessary.

  Approximate Models: Memory in the Real World

  Conscious thinking takes time and mental resources. Well-learned skills bypass the need for conscious oversight and control: conscious control is only required for initial learning and for dealing with unexpected situations. Continual practice automates the action cycle, minimizing the amount of conscious thinking and problem-solving required to act. Most expert, skilled behavior works this way, whether it is playing tennis or a musical instrument, or doing mathematics and science. Experts minim
ize the need for conscious reasoning. Philosopher and mathematician Alfred North Whitehead stated this principle over a century ago:

  It is a profoundly erroneous truism, repeated by all copy-books and by eminent people when they are making speeches, that we should cultivate the habit of thinking of what we are doing. The precise opposite is the case. Civilization advances by extending the number of important operations which we can perform without thinking about them. (Alfred North Whitehead, 1911.)

  One way to simplify thought is to use simplified models, approximations to the true underlying state of affairs. Science deals in truth, practice deals with approximations. Practitioners don’t need truth: they need results relatively quickly that, although inaccurate, are “good enough” for the purpose to which they will be applied. Consider these examples:

  EXAMPLE 1: CONVERTING TEMPERATURES BETWEEN FAHRENHEIT AND CELSIUS

  It is now 55°F outside my home in California. What temperature is it in Celsius? Quick, do it in your head without using any technology: What is the answer?

  I am sure all of you remember the conversion equation:

  °C = (°F–32) × 5 / 9

  Plug in 55 for °F, and ºC = (55–32) × 5 / 9 = 12.8°. But most people can’t do this without pencil and paper because there are too many intermediate numbers to maintain in STM.

 

‹ Prev