Human-Human Conversation vs. Human-Computer Conversation

Provided that features of human-computer interaction are comparable to human-human interaction (which will be called HHI), it should be possible to map the features of HHI to those of HCI. Sacks et al. introduced a model of turn-taking organization which is context-free on the one hand, and capable of extraordinary context-sensivity on the other (Sacks et al. 699). The features discussed by Sacks et al. are:

1. Speaker-change recurs, or at least occurs. (700)
2. Overwhelmingly, one party talks at a time.
3. Occurrences of more than one speaker at a time are common, but brief.
4. Transitions (from one turn to a next) with no gap and no overlap are common. Together with transitions characterized by slight gap or slight overlap, they make up the vast majority of transitions. (700-701)
5. Turn order is not fixed, but varies. (701)
6. Turn size is not fixed, but varies.
7. Length of conversation is not specified in advance.
8. What parties say is not specified in advance.
9. Relative distribution of turns is not specified in advance.
10. Number of parties can vary.
11. Talk can be continuous or discontinuous.
12. Turn-allocation techniques are obviously used. A current speaker may select a next speaker (as when he addresses a question to another party); or parties may self-select in starting to talk.
13. Various 'turn-constructional units' are employed; e.g. turns can be projectedly 'one word long', or they can be sentential in length.
14. Repair mechanisms exist for dealing with turn-taking errors and violations; e.g., if two parties find themselves talking at the same time, one of them will stop prematurely, thus repairing the trouble.

Regarding the first point, it is assumable that every input made by a user invokes a response by the system. While Brennan considers the feedback of a system analogous to backchannels (395), it could also be regarded as an answer to the user's input. Even if a user turns a computer on and off, the system has "answers" to the "input".⁷⁴ Thus, the first feature of HHI seems to be present in HCI.
The second point seems to be different: since most systems are multitasking systems, several programs may make utterances at one time (which shows that the seventh feature is present in HCI).⁷⁵ Dialog boxes do not always appear on top of all windows which a user might have opened, and thus the user might not even know that several parties are speaking (provided that each program is regarded as a conversation partner as pointed out in section 5.1.7). Consequently, the third and the fourth feature pointed out by Sacks et al. seem to be different in HCI, too.
If, however, only the program with which the user works at a given time is regarded as the main conversation partner, the second and the third features might also be true for HCI as well. Nevertheless, empirical data is necessary in order to examine these features.
Turn order, size, and distribution (fifth, sixth, and ninth feature) are not fixed as long as the CPU scheduling algorithm is not regarded as the local management system. While the seventh feature is present in HCI, the eight feature is not, as pointed out in section 4. The eleventh feature, continuous and discontinuous talk, is present in HCI, but discontinuous talk in HHI is different from those in HCI. While a silence in a human-human conversation may be embarrassing, it is not embarrassing at all if a user lets a system wait. A system does not care about the time frame in which an answer is given. On the other side of the coin, it is unacceptable to users in most cases if the system lets the user wait (see section 3.3.6).
Regarding the twelfth feature, turn-allocation techniques in HCI are not the same as in HHI. Modal dialog boxes require the user to give an answer before he can continue his work; other dialog boxes such as the one shown in Figure 4 do not prevent the user from doing something else (and thus making another utterance). However, if the user wants to make an utterance to the application which displays the dialog box (Windows Explorer in this case), the user has to wait until the system's utterance has ended (until the progress bar has reached the end). Mention has to be made that most systems interrupt the user whenever they want (see Figure 15 for a rather polite interruption which is an exception), whereas users might be more polite. However, novice users in particular might not recognize that a system is busy and not capable of processing input.⁷⁶
As pointed out in previous sections, several turn-constructional units are present in HCI; a system may only play a sound in order to signal an error, but it may also display a dialog box with detailed information about the error. Similarly, users might select a menu item (one word) on the one hand or perform a sequence of actions such as dragging an icon onto another icon (several words).
Finally, Sacks et al. state that human conversation features repair mechanisms. While users might stop (or are even forced to stop their work due to a modal dialog box) whenever the system makes an utterance, a system will not always stop whenever it detects user input. Although user input should be given the highest priority, this is not always possible, for example when performing real-time calculations. Similarly, a conversation partner might continue to speak when she regards her utterance as very important.

 

Figure 15: A Qualcomm Eudora Dialog Box

Since Sacks et al. made an attempt to describe features which can be found in any conversation, some differences between human-human conversation and human-computer interaction are not presented. As an example, small talk and meta-movers cannot be found in human-computer interaction. On the other hand, human-computer interaction includes features which are not present in human-human conversation: in most systems, a user action can be undone; some programs like Adobe Photoshop even allow more than one step to be undone. Such foregiveness cannot be found in human-human communication.