All in due time: The development of trust in distributed groups

Jeanne WilsonThe College of William & Mary

School of Business Administration

March 17. 2003

Trust in Distributed Groups

Overall research program objective: Understand interpersonal relations (trust) at a distance

Motivation• Paradox of trust in distributed groups• Competing theoretical explanations• Inadequacy of existing theory

Distributed Groups

Groups in which some or all of the members do not work in the same physical location

• 52% of large companies use geographically distributed teams (deLisser, 1999)

• Collaborative work in a virtual arrangement has been cited as a top workforce trend in the next 10 years (Kemske, 1998)

TrustTrust - willingness to be vulnerable based on

positive expectations of the intentions or behavior of others (Mayer, Davis & Schoorman, 1995; Rousseau, Sitkin, Burt & Camerer, 1998)

Cognitive trust - beliefs about reliability and dependability (McAllister, 1995)

Affective trust - beliefs about reciprocated care and concern (McAllister, 1995)

Overall program planStudy 1 - Lab experiment, test of existing

theory (Wilson, Straus & McEvily, 1999)

Study 2 - Field study, focuses on distributed groups in context, develops a broader theory of interpersonal relations at a distance

Common denominators: looking at trust over time in distributed groups

Study 1: Competing theoretical perspectives

Cues Filtered Out - Computer-mediated communication reduces social context cues and leads to depersonalization (Kiesler, Siegel & McGuire, 1984; Daft, Lengel & Trevino, 1987)

Social-identity Deindividuation - Group identity is the most salient cue computer-mediated groups have; this leads to social self-categorization (Lea & Spears, 1992)

Social Information Processing - All groups are motivated to develop social relationships. It takes longer in computer-mediated groups because there is less social information per message (Walther, 1992; 1995)

0

50

100

0

50

100

FtF CMC

0

50

100

Study 1: Sample and Task

• 52, 3-person groups (participants randomly assigned to group, groups randomly assigned to condition)

• Each group met three times, with the following cycle of tasks: narrow down a list of stocks to 3 that members would research (together or separately) spend tokens on researched stocks (cooperating or defecting)

Four conditions

FFF

EEE

EFF

FEE

Cognitive Trust

The interaction of Condition X Time was significant (F6,98 = 3.69, p < .01).

3.25

3.75

4.25

4.75

Time1 Time2 Time 3

Cog

nitiv

e T

rust

EEEEFFFEEFFF

Affective Trust

The interaction of Condition X Time was significant (F6,98 = 3.09, p < .01).

3.5

4

4.5

5

Time1 Time2 Time3

Affe

ctiv

e T

rust

Sco

re

EEEEFFFEEFFF

Time by condition interaction was significant using Generalized Estimating Equations - for categorical variables over time (B = .90, Z = 3.95, p < .0001).

Cooperation

0

0.5

1

Time1 Time2 Time3

Coo

pera

tion

EEEEFFFEEFFF

Condition effect was significant (B = 1.18, Z = 2.25, p < .05); Time by Condition interaction was marginally significant (B = .53, Z = 1.73, p < .10)

Reliance

0

0.5

1

Time1 Time2

Rel

ianc

e EEEEFFFEEFFF

Conclusions

• Results support social information processing predictions (trust develops more slowly in computer mediated groups)

• Starting condition matters• Prescriptions for practice depend on the

nature of the group length of time

malleability of the task

Limitations

• Student teams lack a “shadow of the future” or structural assurance (which are likely to affect the development of trust in organizational groups)

Inadequacy of existing theories

• All of the theories about development in distributed groups are about media effects

• Distributed groups differ from co-located groups on more dimensions than the technology they use to communicate Distance Familiarity Face-to-face contact Identity

What we know about distanceThe original law of propinquity (Newcomb, 1956)

Physical proximity Frequency of interaction Similarity

Liking

The Site

• Large bank in the midwest• Corporate assets: $30 billion• Provides trust, investment, and retirement

services• Corporate group transitioning to a new team

structure: some teams co-located, some not

78 Teams Investments Relationship Manager

NTRC Account Manager

Daily Valuation Treasury Consultant

Global Accounts Info. Delivery Analyst

RPS Consultant

Design

Quantitative: XP O X1 X2 X3 X4 X5 X6

O X6

Qualitative: Three teams: varying on distance, familiarity, amount of face-to-face contact

• interviewing all members of the teams once a month regarding expectations, trust, violations, attributions, and other team processes that might be affecting trust.

• attending all (formal) team meetings

What predicts trust between team members at month 3?

VariableModel

1 2 3

Distance(log)

-.01* (.002) -.01* (.002) -.00 (.002)

Face-to-face contact .07 (.051) .07 (.053)

Technology use (phone ande-mail contact)

.03 (.056) .05 (.057)

Familiarity(pre-test)

.08**(.025)

a df = 24. The dependent variable is the mean trust level in each dyad at month 3.Standard errors are in parentheses. * p < .05**p < .01

From HLM analyses:

Extras

Full 3-way longitudinal modelLevel 1 model - dyadic variables over time

(trust and communication)Level 2 model - variation among dyads within

a group (familiarity and distance)Level 3 model - variation between groups

(group identity)

Results from Level 1 analysis

• trust between team members is increasing over time (t = 42.46, p < .001)

• the amount of communication between team members has an effect on the development of trust over time (t = 19.31, p < .01)

Level 2 & 3: Intercepts as outcomes

• familiar dyads do start with higher levels of trust (t = 4.49, p < .001)

• trust between individual team members is marginally higher in groups with higher levels of group identity (t = 2.12, p < .10).

Level 2 & 3: slopes as outcomes

• Group identity does not influence the rate of trust development between individual members of the teams (t = 0.67, ns).

• Dyads who are familiar with each other at the outset have a slower rate of growth in trust development than dyads who are not familiar with each other (t = -2.98, p < .01).

Descriptive Statistics M SD 1 2 3 4 5 6 7 8 9 10

1. Affective trust time 1 4.29 .59 .82

2. Affective trust time 2 4.35 .54 .45** .85

3. Affective trust time 3 4.33 .61 .30** .51** .86

4. Cognitive trust time 1 3.95 .45 .54** .34** .18 .88

5. Cognitive trust time 2 4.14 .45 .35** .71** .41** .45** .84

6. Cognitive trust time 3 4.15 .49 .33** .47** .66** .40** .62** .87

7. Cooperation time 1 .80 .39 .08 .18* .11 .24* .18* .09 --

8. Cooperation time 2 .83 .40 .14 .06 .09 .18 .21* .10 .26* --

9. Cooperation time 3 .69 .46 .14 .16 .32* -.07 .17 .24* .09 .18* --

10. Reliance time 1 .46 .50 .07 .02 -.15 .13 .08 -.15 .06 .26** .07 --

11. Reliance time 2 .46 .49 .08 .07 .14 .09 -.16 -.14 .12 .27** .19 .66**

Means and standard deviations over timeA priori contrasts

Measure

Time 1

M SD

Time 2

M SD

Time 3

M SD T2a

t dT3b

t d

Cognitive Trust EEE

EFF

FEE

FFF

3.65 .51

3.59 .62

4.11 .49

4.07 .50

3.89 .52

4.19 .52

3.97 .44

4.18 .62

4.04 .64

4.06 .68

4.01 .51

4.18 .62

1.86† .49

7.16* .98

1.34 .39

-.88 .23

2.48* .66

.88 .24

.21 .06

.64 .17

Affective Trust EEE

EFF

FEE

FFF

4.20 .64

4.16 .70

4.47 .44

4.34 .53

4.17 .64

4.54 .45

4.26 .49

4.42 .50

4.38 .54

4.42 .61

4.22 .49

4.27 .78

.25 .07

2.81* .78

1.80† .52

-1.14 .23

1.89* .50

.43 .12

-2.14* .64

1.08 .29

A priori contrasts Measure Time 1

M SD

Time 2

M SD

Time 3

M SD T2a

t dT3b

t d

Cooperation EEE

EFF

FEE

FFF

.480 .52

.760 .43

.970 .00

1.00 .00

.720 .46

.820 .37

.820 .48

.970 .00

.640 .49

.830 .37

.650 .48

.630 .51

1.55† .42

.56 .16

-1.17 .35

-1.00 .22

.28 .07

.36 .10

-2.01* .60

-2.59* .74

Reliance EEE

EFF

FEE

FFF

.360 .49

.230 .43

.460 .51

.790 .42

.460 .49

.460 .51

.410 .48

.500 .52

-- --

-- --

-- --

-- --

1.75† .46

1.15 .32

-1.48 .40

-1.25 .32

-- --

-- --

-- --

-- --