Introduction to User Studies

Three Types of Studies

1. Traditional studies of users and usage
(often of libraries)
2. Studies that include Internet uses and
similar measures
3. Studies of Digital Libraries and their users
(“digital communities,” just starting)
 Traditional Studies
Focus of
the study

Studies of Studies
Users of Usage

- socio-economic
factors - transaction
-ethnicity, national data, both
origin, language, automated and
etc. traditional
- non-users
 Basic Problems
• Is there a theoretical basis to the analysis?
• Are the results reproducible?
• What can be changed based on the
findings?
• How do we know changes will help?

Possible solution:
- Develop theoretical approach early
Typical Steps in Social Science Research

•State general nature of problem, question, relationship

•Review previous studies
•Create and state “stylized facts”
•Develop testable hypotheses
•Discover or develop data to be used to test hypotheses
•Perform appropriate (statistical) tests
•Describe results
•Draw conclusions
Traditional Bibliometric Model of Scholarly Productivity

Socio-economic factors
Institutional factors
Funding Research output
Revised Bibliometric Model of Scholarly Productivity
Socio-economic factors
Institutional factors

Funding Research output

Computer skills/literacy

Computer use /
Internet use
Revised Bibliometric Model of Scholarly Productivity
Socio-economic factors
Institutional factors

Funding Research output

Computer skills/literacy

Computer use /
Internet use

Major issues:
- What sample/universe to use?
- How to measure inputs?
- How to measure outputs?
- Functional form? (linear?, additive?)
 List of Selected Processes

Ma i l Pn e ws a r c h i e c n r me n u e l m e u d o r a f i n ge r f t p gl a d i s g o ph e r
i ma p i r c k e r mi t l y n x ma i l ma i l t o o l ma i l x me l v yl n e t s c a p e n e wma i l
nn ns l o ok u p pi ne po p p e r r l ogi n t al k t e l ne t tin trn t r n- a r t c
we a t h e r we b s t e r x r n r s h r c p who i s
 Percentage of Faculty Using Each
Service

100 94%
90
80
70 62%
60
% of Faculty

50 44%
37% 35%
40
25% 24%
30
16% 13%
20
9%
10 3%
0

FTP

WAIS
WWW

Gopher

Sequence
Telnet

Newsgroup
E-mail

Supercomp
Listserver

E-journal
ana.

uter
Services
 % of Faculty Using Internet Services
from Questionnaire and Log

100
90
% of faculty using the

80
70
60 Que %
service

50
40 Log %
30
20
10
0 Gopher

FTP
WWW
Telnet

Newsgroups
E-mail

Services
Our main hypothesis in this research is that scholars'
Internet-use data add explanatory power to models of
scholarly productivity. The formal null and alternate
hypotheses are:
H0: Internet use data does not add explanatory
power to the Traditional Publication Model.
H1: Internet use data adds explanatory power to
the Traditional Publication Model.
We used multiple regressions to estimate the four models
listed in Figure 5. Since we are interested in whether a set
of one or more Internet-use variables “improves” the
explanatory power of the traditional model, the appropriate
measure is the F-statistic from a comparison of the
restricted (traditional) model with the unrestricted (new)
model.
Dependent Variable: PUBAV

Analysis of Variance

Sum of Mean
Source DF Squares Square F Value Prob>F
Model 6 56.72845 9.45474 1.230 0.3148
Error 35 269.07034 7.68772
Total 41 325.79879

Root MSE 2.77267 R-square 0.1741

Dep Mean 3.78048 Adj R-sq 0.0325
C.V. 73.34194

Parameter Estimates

Variable DF Parameter Standard T for H0: Prob > |T|

Estimate Error Parameter=0
INTERCEP 1 22.344070 8.16887630 2.735 0.0097
AGE 1 -0.615779 0.24909898 -2.472 0.0184
PHD_AGE 1 0.632716 0.25079432 2.523 0.0163
TIMTOPHD 1 0.460290 0.33084979 1.391 0.1729
RL 1 -0.020384 0.02346329 -0.869 0.3909
CAR_SCR 1 -0.597347 0.54715391 -1.092 0.2824
PIM1 1 0.051832 0.28491924 0.182 0.8567

F ig u re 4: Th e b e s t t ra d itio n a l m o d e l o f p u blic a tio n

Mode l Va r i a bl e s :

Ra w I nt e r ne t Da t a :

1 AGE PHD_AGE RL PI M1 COMP1 COMP2

2 AGE PHD_AGE RL PI M2 COMP1 COMP2

Tr a ns f or me d I nt e r ne t Da t a :

3 AGE PHD_AGE RL PI M1 COMP1 COMP2

4 AGE PHD_AGE RL PI M2 COMP1 COMP2

F ig u re 5: P u blic a t io n m o d e ls w it h In te rn e t p rin c ip a l c o m p o n e n t s
Eqn. MODEL I nt e r ne t N R2 r R2 ur F CV ( 5 %)
Da t a

1 AGE, PHD_AGE, RL, PI M1 , Ra w 23 0. 1765 0. 4811 4. 04 3. 63

COMP1 , COMP2

2 AGE, PHD_AGE, RL, PI M2 , Ra w 23 0. 1808 0. 4806 4. 02 3. 63

COMP1 , COMP2

3 AGE, PHD_AGE, RL, PI M1 , Tr a n s f o r m 23 0. 219 0. 5662 6. 4 3. 63

COMP1 , COMP2

4 AGE, PHD_AGE, RL, PI M2 , Tr a n s f o r m 23 0. 2199 0. 5663 12. 77 3. 63

COMP1 , COMP2

KE Y:

N Nu mb e r o f c a s e s

R2 r R2 o f t h e r e s t r i c t e d mo d e l

R2 u r R2 o f t h e u n r e s t r i c t e d mo d e l

F Th e v a l u e o f t h e F s t a t i s t i c

CV Th e c r i t i c a l v a l u e o f t h e F d i s t r i b u t i o n

F ig u re 6 : Te s t s o f Hy p o t h e s e s —Mo d e ls a n d Re s u lt s
Variable Eqn. 1 Eqn. 2 Eqn. 3 Eqn. 4

Constant 17.1 17.2 11.2 11.2

Age -0.313 -0.317 -0.216 -0.215

Ph.D.-age 0.197 0.201 0.138 0.138

RL 0.00860 0.00777 0.00563 0.00565

PIM -0.0794 -0.051 0.0515 0.0392

COMP1 (raw) 19.6 * 19.7 *

COMP2 (raw) 2.99 * 2.99 *

COMP1 (trns) 1.84 * 1.84 *

COMP2 (trns) 0.328 0.324

R
2 0.481 0.481 0.566 0.566

NOTE: * = st a t ist ica lly sign ifica n t ; p<0.05.

F ig u re 7: Re g re s s io n Re s u lt s (In t e rn e t U s e Equ a t io n s )
The exact equation, as estimated in Equation 1, is:

PUBAV = 17.1 – 0.313 AGE + 0.197 PHD-AGE + 0.0860 RL – 0.0794 PIM1

+ 19.6 (Login-FTP principal component)

+ 2.99 (Kermit principal component).

 Std. Regression Effect on
Variable Mean Std. Dev. Mean + 10% Z-score Scoring Coeff. PUBAV
Coeff.

ALLL 20.24 94.81 22.27 0.0214 0.4939 19.65 0.207

FTP 160.1 609.4 176.2 0.0263 0.4927 19.65 0.254

KERMIT 9.643 31.99 10.61 0.0301 0.4499 2.992 0.041

F ig u re 8: Effe c ts o f In te rn e t-U s e Va ria b le s

Variable Mean Coefficient 10% of Effect on Elasticity of

mean PUBAV Substitution (*)

AGE 52.4 -0.313 5.24 -1.6385 -7.91

PHD-AGE 23.0 0.197 2.3 0.4523 2.18

RL 57.6 0.00860 5.76 0.0495 0.24

PIM1 4.17 -0.0794 0.417 -0.0331 -0.16

NOTE : * ca lcu la t ed r ela t ive t o ch a n ge in ALLL

F ig u re 9: E ffe c t s o f Ot h e r Va ria ble s a n d E la s tic itie s

 Digital Libraries
and Collaborative Knowledge
Construction
Based on a presentation by
Nancy Van House
September 15, 1999
 Digital Libraries
• Support knowledge work
• Knowledge work occurs in material and
social matrices
• DLs mutually constituted with work,
practices, tools, artifacts, knowledge,
knowledge communities
 Evaluating DLs
• Direct effects: how well they meet users’
needs & expectations
• Higher-order effects: Understanding how
DLs support and potentially change
– Work
– Its practices and artifacts
– People/communities who do the work
– Institutions that support it
 Possible Theoretical Bases for
DL Evaluation
• Situated action
– work and learning
• Science studies
– knowledge creation
– knowledge communities
• Actor-Network Theory
– knowledge creation and communities
– stabilization of socio-technical systems
 Characteristics of Knowledge
Work
• Situated in place and time
• Distributed x people, places, time
• Social - collectively decide what is known,
how we determine what is known, how we
do the work
The UC Berkeley Digital Library
Project
• Supporting environmental planning
• Innovative tools
• Large, diverse testbed
 This Research

• To better understand distributed,

collaborative cognitive work, the role of
information and information practices and
artifacts, and the potential effects of digital
information
• To investigated the practices of
environmental planning, particularly
information use and production
 This Study
• Not about the UC Berkeley DL but about
potential of DLs in general, and current
issues around sharing of digital data.
• Method: Interviews
 Knowledge Communities
• Two engaged in work related to environmental
planning
– Water planning
– Botanical
• Commonalities
– Multiple scientific (mostly) disciplines
– Various organizational settings
– Large datasets of observations and summaries and
analyses
• collected by many, over time, differing methods
– Potentially major economic impact
– Often highly political
 Findings
• Contributions
• Use
• Cooperation in creating and operating DL
 Findings: Barriers to
Contribution
• Misuse of data
– People who ‘don’t understand’
– “Using our data against us’
• ‘Productizing’
– Burden
– Mismatched incentives
– Making invisible work visible
 Findings: Use
• User must be willing to trust both content and
functionality of DL
• Assessing trustability of contents
– Assessing sources
• Ability and training
• Interests
– DL’s filtering policies and processes
• DL’s functionality
– E.g., search tools, GIS overlays
 Findings: Cooperation in
Creating and Operating DL
• Both a research project and active testbed
• Groups to interest: computer scientists and
applications area specialists
• Cutting edge vs reliable, incremental improvement
over past
• Differences in decision-making styles:
hierarchical, planned vs distributed, emergent
• Articulation work
 Possible Theoretical Bases for
DL Evaluation
• Situated action
– work and learning
• Science studies
– knowledge creation
– knowledge communities
• Actor-Network Theory
– knowledge creation and communities
– stabilization of sociotechnical systems
Two Broad Approaches to Social
Theory
• Reductionist: economics, behavioral
psychology, structural-functional sociology
• Irreductionist: ethnomethodology
(Garfinkel), practice/structuration theory
(Giddens, Bourdieu, Lave)
 Irreductionist Approaches
• Social order/institutions not pre-given but
continually recreated through the concrete, day-
to-day activities Interest in processes by which
people construct meaning and re/produce social
order
• Interest in processes by which this done
• Emphasis on practices and artifacts which shape
understanding and carry it across time and space
 Situated Action, Situated
Learning
• Knowledge not inert substance transferred
from teacher to student
• Knowledge and learning on-going
construction of meaning through action and
interaction
• Knowing not activity of single mind but
complex social and material phenomenon,
“nexus of relations between the mind and
work and the work in which it works”
(Lave, 1988).
 Communities of Practice: the Social
Matrix of Knowledge Work
• People learn & work within groups who share
understanding, practices, technology, artifacts, and
language, e.g., professions, workgroups, and
disciplines.
• A person typically belongs to multiple
communities of practice
• Important task of communities of practice:
– deciding what is known;
– the processes and principles by which
knowledge claims are evaluated;
– who is entitled to participate in the discussion;
– whom to believe.
 Information Artifacts
• Include texts, images, maps, databases,
thesauri, classification systems…
• Not simply reflections of knowledge but
instrumental in creating it, coordinating
work across space and time
• Tightly bound up with practices and
communities
 Science and Technology Studies
• Science Studies
– practices of scientific work
– how the participants come to an agreement about what counts as
fact or discovery
– what inferences are made from facts
– what is regarded as rational or proper conduct
– how credibility of claims is assessed
• Science and Technology Studies (STS)
– Increased emphasis on technology, esp’ly large sociotechnical
systems
• Science as prototype of rational knowing > applicable to
knowledge work generally
 Approaches to Research
• Sci & tech knowledge neither purely natural nor
social
• Attention to processes of sci knowledge production
• Methodological orientation
– “Follow the actor”
• Attention to activities
• Not deciding ahead of time where to draw boundaries
– Funding as part of scientific activity
– Inscriptions play key role in sci knowledge production
• E.g., labs as producing papers
• Carrying work across space and time
 Actor-Network Theory
(Bruno Latour, Michel Callon, John Law)
• Not a theory but a “method of looking at the actors’ own
world building activities”
• Doesn’t take the stability of sociotechnical systems as
given, but asks how they are created and maintained.
• “Organization is an achievement, a process...Its
components -- hierarchies, organizational arrangements,
power relations, flows of information -- are uncertain
consequences of the ordering of heterogeneous materials..
ANT is thus a theory of agency, of knowledge, and of
machines.” (Law in systems science (Emphasis added)).
Key Concepts of ANT
• Action-at-a-distance
• The actor-network
• Translation
• Black-boxing
• Intermediaries
 Key Concepts of ANT
• Action-at-a-distance is the problem of coordinating
and sustaining collective activities over space and
time).
• The actor-network “is most simply defined as any
collection of human, non-human, and hybrid
human/non-human actors who jointly participate in
some organized (and identifiable) collective activity
in some fashion for some period of time.”
– Perhaps the most radical contribution of ANT is the
inclusion of the non-human in its “heterogeneous
networks.”
 Key Concepts, cont.
• Translation is a key process by which actor-
networks are created and stabilized, however
temporarily. Actors’ disparate interests get translated
into a set of interests that coincide in the network.
• Black-boxing is a process of closing questions and
debates.
• An intermediary is an actor (of any type) that stands
between two others and translates between them in
such a way that their interaction can be more
effectively articulated.
– Inscriptions, Latour’s “immutable, combinable mobiles:”
 Approaches to Research
of Situated Action, Science
Studies, ANT
• Contention that sci & tech knowledge
neither purely nature nor purely social
• Methodological orientation:
– “Follow the actor” (ethnography)
– Inscriptions
 Applied to our DL
• Applications area: environmental planning
• The DL as a large sociotechnical systems
• Method
Implications: Joining Findings & Theory
I. DLs and Knowledge Productions

• DLs support knowledge work, which is

situated, distributed, and social
• DL ‘evaluated’ according to its relationship
with knowledge work and its social and
material matrix
Users’ and Providers’ Concerns – the DL
and Knowledge Production
• Fear of misuse of info, inability to evaluate
credibility = crossing boundaries of communities
and assemblages within which/for which created
• Practices of (1) work and (2) assessing/demo’ing
credibility undermined by
– Ready crossing of sociotechnical boundaries
– Fluid recombination of inscriptions
– Opening of black boxes, e.g. document, methods of
work
– Closing of others, e.g. analytical and search tools
Implications: Joining Findings & Theory
II. The DL as a Sociotechnical System

• Successful DL enrolls dynamic network of

users, sources (human and nonhuman),
builders, and technology
• Requires on-going enrollment and multiple
translations
• Stabilization not equilibrium but on-going
achievement
 Implications for DL Design
• Articulate with existing assemblage of practices,
artifacts, participants
– Attention to existing and emergent practices of trust
and credibility
• Design specific to communities and tasks
– Tension of specificity and generality
• Include high-order user involvement
• Design to be emergent and dynamic – on-going
enrollment and co-constitution
 Implications for DL Evaluation
• Not just how well it meets users’ needs but
how co-constituted with assemblages of
work
• Target to specific user communities and
tasks
• Consider not just service delivery but
stabilization of DL as a heterogeneous
network
 Evaluation Methods
• Ethnographic
– “Follow the actor”
– Study activities in natural settings in which they
occur
– Focus on what people actually do, not simply
their accounts of behavior
• Varied
– Different methods appropriate for different
issues
 Conclusion
• Understanding knowledge work not simple
• Beginning to have an analytical approach from
which to better understand KW and DLs
• Applies to other areas of electronic collection,
delivery, manipulation of information where trust
and credibility are important
– E.g., knowledge management