Nancy Griffeth
Bellcore MRE 2Q-382,
445 South Street, Morristown NJ 07960
Without features, the telephone interface is simple
and can easily be taught to a child, but the proliferation of
telephone features makes the telephone's behavior hard for an
adult to understand. The difficulty arises not just from the need
to learn several ways of using the telephone, but also from the
interactions among features, which can cause each feature to behave
differently in the presence of other features. In this paper,
we describe some of the problems feature interactions present
to telephone users, and discuss their extent.
Two or more features of a telephone service are said to interact if one changes the functionality of the others. This creates a problem: a telephone user must learn the interactions of each collection of features in addition to the individual feature behaviors. There is a growing body of work describing the related software development problem and presenting possible solutions for today's telephone network [1,2,3,4]. However, there has been no work on the human aspect of this problem.
The problem would seem to worsen as more 'intelligence'' is introduced into telephones and their features multiply. In the near future, communication will increasingly take place over broadband networks, using mechanisms for video and data communication as well as voice. New services will offer more flexible interfaces, which will reduce the likelihood of feature interactions, counter-balanced by more functionality, which will increase the likelihood of feature interactions.
Software companies seeking to differentiate their products will inevitably be tempted to add features, and we may even be able to download new features dynamically, as is done with Java on the World-Wide Web. In such an environment, will users be able to cope with a proliferation of features?
The goal of this paper is to document the problem as it exists in the telephone system today, to discuss its extent, and to present a challenge to researchers in human interfaces.
Donald Norman pointed out on page 188 of ``Design of Everyday Things'' [5] that a good design will have two properties:
(1) the user can figure out what to do, and
(2) the user can tell what is going on.
When new features are added to a device, feature interactions can interfere with these properties, even if the original device and all its features are individually well-designed.
This section illustrates this point with examples
of feature interactions on a telephone. It is interesting to note
that Norman presents the ordinary, unfeatured telephone as an
example of good design, and presents several cases of telephones
with numerous features as examples as bad design.
A telephone provides a limited number of signals. The same signals are used to mean different things in different contexts. Consequently, it can be difficult for a user to determine what to do in order to accomplish a desired goal. For example, the ''flash-hook'' or ''flash'' signal is used by both Three Way Calling and Call Waiting.
Three-Way Calling. Suppose that Nancy wants to make a conference call including herself, Greg, and John. First, Nancy calls Greg. After Greg answers, she uses ''flash'' to make the second call. After the flash, Greg is on hold and she hears dial-tone. She then calls John. When John answers, she uses ''flash'' again to join the three parties in a single call. Subsequently, if she uses ''flash'', John will be dropped from the call.
Call Waiting. Call Waiting is a popular feature that permits the Call Waiting subscriber to answer the phone even when talking to someone else. Suppose that Nancy is talking to Greg, and John calls. If Nancy subscribes to Call Waiting, instead of John hearing the busy tone, he hears ringing and Nancy hears a special tone (interrupting her conversation with Greg briefly) that indicates a call. She can then use the ''flash'' to talk to John. While she is talking to John, Greg is on hold. Subsequent uses of flash swap the held and active calls, so that she can switch back and forth between them.
In the first example, Nancy probably wants Call Waiting to continue working while she is on the Three-Way Call with Greg and John. Also, in the second example, she may well want to join the separate calls with Greg and John into a single conference call. But neither is possible, because the flash signal is ambiguous: it can mean ''Get dial-tone for a new call'', ''Add a new party to the current call'', ''Switch from the active call to the held call'', or ''Drop the last party added'', depending on the context. A telephone user who thinks about the problem may well recognize the ambiguity. How does he determine what can be done and how to do it?
In the current telephone system, the answer is, ``Learn
arbitrary rules''. When this kind of thing happens, the control
software in the network usually resolves the interaction by de-activating
one feature to avoid the ambiguity. The caller must learn that
these features cannot be active at the same time, and the functionality
of each is lost when the other is active.
Another kind of feature interaction occurs because
there are only a few signals with which telephone can signal its
state, and the same signal may mean different things when different
features have affected a call. As a consequence, a user may not
be able to tell what is going on.
I was rather surprised recently to realize that there are not one but three reasons for the telephone to ring:
1. someone (or some computer) is calling;
2. the caller has hung up, but a call is on hold; or
3. a monitored line is now available, and a call can be made to it.
The second reason is one of the most common reasons for the telephone to ring in my household. The sequence of events causing this is that someone calls voice mail to pick up messages, then returns a call to one of the callers. When hanging up voice mail, the switch-hook was depressed only briefly; the control software thinks the signal was ''flash''; and voice mail winds up on hold. The voice mail remains on hold throughout the subsequent call, and when the caller hangs up, the voice mail rings us back -- with the surprising effect that when we pick up the telephone, we hear our voice mail menus!
This has an unfortunate side-effect for features
like ''Return Call'', which returns a call to the last caller.
Since only the first case represents a real call, Return Call
is appropriate only in that case; in the other two cases, it is
inappropriate.
Bellcore specifies required interactions among features that existed prior to AT&T's divestiture. Switching system vendors also specify the interactions among features implemented on their products, and Bellcore documents some current interactions and potential interactions among more recently-developed features [6].
Recently, I examined 8 common features (Automatic Recall, Call Waiting, Call Forwarding Busy Line, Call Pick-Up, Calling Number Delivery, Selective Call Rejection, Three-Way Calling, and 911). There are 28 ways to pair 8 features. Twenty of the possible 28 pairs in this collection of features had specified interactions. There were a total of 37 identified interactions (some pairs had more than one interaction).
The challenge for human factors research is to determine
how one resolves interactions in a manner natural to users of
a communications device. Even recognizing that only user studies
can make the ultimate determination of design quality, a collection
of principles that one can follow to develop possible resolutions
would represent a major advance.