Page 75 -

44 Chapter 2 Understanding and conceptualizing interaction

research has been carried out, however, on the best way to design the ordering and
sequencing of button pressing for physical devices like cell phones, calculators, re-
mote controls and vending machines.

Another ubiquitous vending machine is the ticket machine. Typically, a number of instruc-
tions have to be given in a sequence when using one of these. Consider ticket machines de-
signed to issue train tickets at railway stations-how often have you (or the person in front
of you) struggled to work out how to purchase a ticket and made a mistake? How many in-
structions have to be given? What order are they given in? Is it logical or arbitrary? Could
the interaction have been designed any differently to make it more obvious to people how to
issue instructions to the machine to get the desired train ticket?

Comment Ticketing machines vary enormously from country to country and from application to appli-
cation. There seems to be little attempt to standardize. Therefore, a person's knowledge of
the Eurostar ticketing machine will not be very useful when buying a ticket for the Sydney
Monorail or cinema tickets for the Odeon. Sometimes the interaction has been designed to
get you to specify the type of ticket first (e.g. adult, child), the kind of ticket (e.g. single, re-
turn, special saver), then the destination, and finally to insert their money. Others require
that the user insert a credit card first, before selecting the destination and the type of ticket.

2. Conversing

This conceptual model is based on the idea of a person conversing with a system,
where the system acts as a dialog partner. In particular, the system is designed to
respond in a way another human being might when having a conversation with
someone else. It differs from the previous category of instructing in being intended
to reflect a more two-way communication process, where the system acts more like
a partner than a machine that simply obeys orders. This kind of conceptual model
has been found to be most useful for applications in which the user needs to find
out specific kinds of information or wants to discuss issues. Examples include advi-
sory systems, help facilities, and search engines. The proposed tourist application
described earlier would fit into this category.
The kinds of conversation that are supported range from simple voice-recognition
menu-driven systems that are interacted with via phones to more complex natural-lan-
guage-based systems that involve the system parsing and responding to user queries
typed in by the user. Examples of the former include banking, ticket booking, and
train time inquiries, where the user talks to the system in single-word phrases (e.g.,
yes, no, three) in response to prompts from the system. Examples of the latter include
search engines and help systems, where the user types in a specific query (e.g., how do
I change the margin widths?) to which the system responds by giving various answers.
A main benefit of a conceptual model based on holding a conversation is that it
allows people, especially novices, to interact with a system in a way they are already
familiar with. For example, the search engine "Ask Jeeves for Kids!" allows chil-
dren to ask a question in a way they would when asking their teachers or parents-
rather than making them reformulate their question in terms of key words and
Boolean logic. A disadvantage of this approach, however, is the misunderstandings
that can arise when the search engine is unable to answer the child's question in the