[PDF] - Elicitation C.W. Johnson, Univ ersit y of Glasgo w, Glasgo PDF Document

SLIDE 1 Elicitation C.W. Johnson, Univ ersit y

f

Glasgo w, Glasgo w, G12 8QQ. Scotland. johnson@dcs.gla.ac.uk, h ttp://www.dcs.gla .a c.uk/ johnso n Octob er 2001 CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001 1

SLIDE 2 Intro duction

P

a rticipato ry and User Centred Design

Interviews,

Questionnaires, F

cus

Groups.

T

ask Analysis. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 3 Recap

Lecture

1

n

mental mo dels.

User's

mental mo del

f

the system:

everything

they lea rned from using it;

t

ypically , no idea

f

the implementation.

Designer's

mental mo del

f

the system:

can

b e igno rant

f

the application domain;

but

do es understand implementation.

Can

designers ever understand users? CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 4 What do es the Company W ant?

Managers

have clea r ideas:

`new

system will reduce stang';

`sta

will p erfo rm task X in N seconds....'

Managers

have no idea:

ma

y not understand everyda y p roblems

f

their sta;

ma

y urge gratuitous use

f

pa rticula r technologies;

ma

y disagree amongst themselves. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 5 What do the Users W ant?

Users

have little idea

f

what can b e built:

the

easy things can b e very ha rd to imagine;

the

computationally intractable can seem easy . Ac kno wledgemen t: BBC

Users
ften

nd it ha rd to sa y what they do:

ver

time p eople build up exp ertise;

skills

b ecome unconscious;

long-term

not sho rt-term memo ry . CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 6 User-Centr e d Design

Solution

involve users in the development p ro cess:

use
f

rapid incremental p rotot yping;

this

supp

rts

fo rmative evaluation;

user

testing b efo re making design decisions.

Problems:
if

tests fail is it the design

r

unrep resentative users?

if

a test fails ho w do y

u

generate new solutions? CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 7 Particip atory Design

Involve

users in the development p ro cess:

mo

re direct involvement than in UCD.

Users

p resent in design meetings:

ho

w rep resentative is the user involved?

ften

rep resentatives resemble develop ers!

Problems:
very

dicult to manage:

can

feel intimidated b y development team;

can

feel sup erio r to the development team. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 8 Ma rk et Surveys

Assess

users' lifest yles and aspirations:

Ea

rly adopters? Second-w ave? Conservatives? Ac kno wledgemen t: BBC

What

comp etito r systems a re successful no w?

Photo

dia ries and technology trials. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 9 Interviews

Unscripted

sets

f

questions:

follo

w the mo

d
f

the interview but...

can

b e (mis)led b y the interview ee.

Prescripted

sets

f

questions:

ask

every

ne

the same things and compa re answ ers;

might

not ask the

ne

really crucial question?

Ho

w to reco rd the resp

nses:
notes

(cheap but ma y detract from interview);

audio

taping (unintrusive but costly to transcrib e);

video

(facial exp ressions but costly to transcrib e). CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 10 Interviews

Questions

can give a w a y info rmation.

Implied

criticism:

`Most
f

the sta here have never used an IX3205B?'

P
r

p repa ration:

Interview

er: `Y

u

w

rk

in the paint p repa rtion shop?'

Interview

ee: `No, that w as closed last y ea r...'

Arrogance:
`W

e're planning a servlet-based extension to the...'

P

atronising intro ductions: `No w please relax, there really is nothing to b e w

rried

ab

ut...'

CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 11 Questionnaires

Lo

w resp

nse

rates esp ecially from certain groups? Do you like the existing system? Yes No

Multiple

choice:

users

ma y mindlessly tick certain resp

nses;
ma

y ask trick questions (double negatives etc). What is good about the existing system?

Op

en ended:

users

ma y not have time/interest to write much;

ho

w t ypical a re the users who write lots and lots? CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 12 Questionnaires

Asking

mo re subtle questions... It is easy to make mistakes? Disagree 1 2 3 Agree terrible wonderful Were the lectures helpful? 1 2 3 4 5 6 N/A

Scala

r resp

nses

(Lik ert scales):

can

b e dicult fo r users to cho

se

b et w een values. Rank the following from 1 to 3 The most important thing about this lecture is:

---that

it finishes in an hour

---that

it helps me to pass the exam

---that

it never mentions Java

Rank

ed resp

nses:
what

if there a re equal p references? CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 13 F

cus

Groups

Questionnaires
p

en to bias and inuence.

P

eer groups complete questionnaires together?

Ma

y miss sha red attitudes and b eliefs.

Ho

w do y

u

interp ret resp

nses?
pa

rtial answ ers ma y need mo re detail;

resp
nses

ma y contradict views

f

managers etc.

F
cus

groups to follo w-up questionnaires. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 14 F

cus

Groups

Y
u

act as facilitato r fo r a discussion: { intro duce a topic and let them talk ab

ut

it; { p rompt and direct to w a rds k ey topics.

Reco

rd and account fo r dierent p ersp ectives.

Problems:

{ can b e `hi-jack ed' b y

pinionated

pa rticipants; { ca re must b e tak en with group dynamics. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 15 P a rticipation Grids

Who

contributed to a fo cus group? { mino r manager talks but p roject sp

nso

r is silent; { might come a w a y b elieving mino r view?

P

a rticipation grids:

dra

w an a rro w when A talks to B;

rep

eat fo r each majo r topic. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 16 Requirements Do cuments

What

do y

u

do with the info rmation?

Construct

a requirements do cument.

Describ

es what must b e done:

p

rovide automated

rdering

facilities fo r all sta;

sta

should complete rst

rder

with 1 da y training.

Do

es not describ e how to do it:

use

a P entium I I I running NT, written in Java...

Can

b e based

n

usage sc enarios. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 17 Hiera rchical T ask Analysis

What

do y

u

do with requirements do cument?

Spilt

a high-level task into sub-tasks. 0. in

rder

to complete an

rder

0.1 take customer's product selection 0.1.1 ask for reference number 0.1.2 enter reference number in system 0.2 take customers contact details 0.2.1 ask for customers post code 0.2.2 enter postcode into system 0.2.3 complete any blank fields 0.3 take customers payment details

Go
d

p

ints:
builds

a conceptual mo del

f

users' view;

can

identify kno wledge requirements with each step.

Simple

plans but what if things go wrong? CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 18 Current Problems: Plans and Situated Actions

Lucy

Suchman criticises much

f

this.

Plans

evolve within a complex w

rking

environment.

P

eople:

a

re mo re

pp
rtunistic;
must

adapt to interruptions.

So

designer must:

consider

users' mo del

f

the system;

consider

the eect that context has

n

the mo del. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 19 Conclusions

P

a rticipato ry and User Centred Design

Interviews,

Questionnaires, F

cus

Groups.

T

ask Analysis. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001

SLIDE 20 F urther Reading

Shneiderman
n:
design

p ro cess

pp.

95-117;

evaluation
pp.

124-150.

He

combines elicitation and evaluation. CS-1Q: HCI (Lecture 3) c

C.W.

Johnson, 2001