cummings02-QualityOfOnlineRelationships.pdf

Q UALITY OF

Online

Social Relationships

O NLINE RELATIONSHIPS ARE LESS VALUABLE

THAN OFFLINE ONES . I NDEED , THEIR

NET BENEFIT DEPENDS ON WHETHER THEY

SUPPLEMENT OR SUBSTITUTE FOR

OFFLINE SOCIAL RELATIONSHIPS .

{ By Jonathon N. Cummings, Brian Butler, and Robert Kraut }

eople use the Internet

intensely for interper-

sonal communication,

sending and receiving

email, contacting

friends and family via

instant messaging ser-

vices, visiting chat rooms,

or subscribing to distribution lists,

among other activities. The evidence is

clear that interpersonal communication

is an important use of the Internet, if not

its most important use. For example,

both self-report surveys [12] and

computer monitoring studies [5] indicate that email

is the most popular online application.

Claims regarding the Internet’s usefulness for

developing social relationships, however, remain con-

troversial. Both personal testimonials (for example,

[10]) and systematically collected data document the

deep and meaningful social relationships people can

cultivate online (for example, [8]).

This evidence, however, conflicts with data com-

paring the value that people place on their online rela-

tionships with offline relationships and with data

comparing social relationships among heavy and light

Internet users. For example, Parks and Roberts [9]

surveyed users of multiplayer environments called

MOOs. Ninety-three percent of the users had made

friends online, but when asked to compare their

online friendships with those offline, respondents

rated offline ones higher. Respondents to Nie’s [7]

national survey reported spending less time with

friends and family since going online, with the decline

greatest among the most frequent Internet users. And

Kraut et al. [6] presented longitudinal evidence to

demonstrate that among new Internet users, online

time diminished social involvement and psychologi-

cal well being.

Understanding the impact of the Internet on

human social relationships requires two types of evi-

dence. First, we need to know how computer-medi-

ated communication affects the quality of particular

social interactions and relationships. Are the online

ones better, the same, or worse than those sustained

by other means? Second, we need to know how com-

puter-mediated communication affects one’s mix of

social interactions and relationships. The impact of

the Internet is likely to be very different if it supple-

ments communication with existing friends and fam-

ily, or if instead it substitutes for more traditional

communication and social ties.

This article addresses the first question by explicitly

comparing online and offline social interaction. We

briefly summarize evidence from several empirical

studies, all of which suggest that computer-mediated

communication, and in particular email, is less valu-

able for building and sustaining close social relation-

COMMUNICATIONS OF THE ACM July 2002/Vol. 45, No. 7

103

T HE

ships than face-to-face contact and telephone conver-

sations. These studies include the following surveys:

3.0

2.5

• International bank employees who describe the

value of particular communication sessions for

work relationships;

• College students, using the same methodology,

but focusing on personal relationships;

• A longitudinal study of new Internet users; and

• Examination of behavior on email-based listservs.

2.0

1.5

1.0

Work

Relationships

Face-to-face

Phone

tion, one might also object

that personal relationships are

not central to work activity,

although many studies stress

their importance for getting

work done. To counter these

objections, we replicated the original study in 1999

among students at an eastern university. These stu-

Comparing Communication Over

Different Media

One way to evaluate the usefulness of the Internet

for developing and maintaining social ties is to ask

people to compare particular communication ses-

sions on relevant outcomes. One can then relate the

outcomes to features of the communication session

(for example, who it was with, the duration, and the

modality over which it occurred). This technique

has been used to uncover features of conversation

that lead to the development of social relationships

in face-to-face settings (for example, [3]). We apply

it to email, telephone, and face-to-face communica-

tion among bank employees and university students.

In our 1991 study, 979 employees of a multina-

tional bank reported on their most recent communi-

cation conducted over different media. About 81%

used email in their jobs, sending an average of 15

messages per week. Respondents evaluated the useful-

ness of communication episodes using criteria related

to the success of work groups, including usefulness for

getting work done and for developing or sustaining a

work relationship, utilizing a 3-point scale, where 1

meant not very useful and 3 meant very useful. We

report data on 5,205 communication episodes that

occurred in person, by telephone, or by email (see [4]

for more detail.)

Figure 1 compares the evaluation of conversations

held over each media for the two measures of success.

Respondents reported communication by email to be

reliably worse than communication conducted face-

to-face or by telephone, both for getting work done

and for sustaining work relationships. However, the

disadvantages of email were significantly greater for

maintaining relationships than for getting work

done. These differences among the media remain

even when one statistically controls for relevant vari-

ables, including respondents’ gender, age, job title,

daily volume of communication, and experience

with email.

One might object that this data comes from the

early years of email, although employees in this firm

had been using email since the mid-1970s. In addi-

Figure 1. Perceived

usefulness of commu-

nication by media and

outcomes in an

international bank.

3.0

2.5

2.0

1.5

1.0

Work

Information

Relationships

Face-to-face

Phone

dents used email extensively,

estimating a mean of 11 mes-

sages per day, and were in a

stage in life that stressed the

importance of developing per-

sonal relationships. Some 39

students completed a diary, recording information

about each of 259 communication episodes in which

they had participated during a four-hour block—late

afternoon to early evening. Students recorded their

relationship with their communication partner (rela-

tive, friend, acquaintance, or other), its duration, the

topic of conversation (schoolwork, personal, or

other), and the modality over which it occurred.

Respondents evaluated each communication for its

usefulness in getting work done, exchanging infor-

mation, and developing or maintaining a personal

relationship. They made their evaluations on 5-point

scale.

Like the banking study, students evaluated email

communication sessions as an inferior means to

maintaining personal relationships compared to those

conducted in person (p <0.05) and by telephone (p <

0.05), these latter being equal (see Figure 2). The stu-

dents, however, found email to be as good as the tele-

phone and in-person communication for completing

Figure 2. Perceived

usefulness of com-

munication by media

and outcome among

college students.

104

July 2002/Vol. 45, No. 7 COMMUNICATIONS OF THE ACM

schoolwork (p > 0.10), and even better for the

exchange of information (p < 0.05).

Students also estimated the frequency of commu-

nication over the different modalities and the strength

of their relationship with each of the 148 partners. We

created an index of relationship strength by averaging

their answers to two questions: “How close do you

feel to this person?” and “How often do you get favors

or advice from this person “(alpha = 0.92). We used

linear regression to predict the strength of the rela-

tionship from frequency of communication with that

partner over the different modalities: email, in-person,

and telephone. Frequency of communication across

all three modalities was significantly related to the

strength of relationship, both directly and once the

partner’s gender, nature of the relations, length of the

relationship, and geographic distance between the

parties were controlled statistically. However, commu-

munication. The first, whom we refer to as the “Inter-

net partner,” was the individual outside of their

household to whom they sent the most email, as

recorded in computer-generated usage logs collected

as part of the project. Some 111 respondents

answered questions about an Internet partner. The

second, whom we refer to as the “nonInternet part-

ner,” was the person outside of their household with

whom respondents claimed to have the most frequent

communication in any modality. Some 125 respon-

dents answered questions about a nonInternet part-

ner. To allow for comparisons between relationships

conducted by email and those conducted primarily

over other modalities, we limit our analyses here to

the 99 respondents who answered questions both

about an Internet and a nonInternet partner, and for

whom these partners were different individuals.

Respondents indicated each partner’s gender and

age, duration of acquaintance, role relation (for

example, family, friend, co-worker), and geographic

proximity (for example, neighborhood, city, state).

Participants then rated their frequency of email, face-

to-face, and telephone communication: (5-daily, 4-

weekly, 3-monthly, 2-less often, 1-never). A 5-point

scale indicated psychological closeness with the part-

ner: “I feel very close,” “I could freely confide in this

person,” “This person is important to me,” and “I

understand this person fully” (alpha = 0.90).

We were interested in three questions: Do people

differ in the overall volume of communication they

have with the people they keep up with using differ-

ent modalities? Do they differ in how close they feel

toward them? Is communication with a partner over

different modalities predictive of differing degrees of

psychological closeness?

The number of respondents’ communication ses-

sions per month, broken and summed over all modal-

ities, indicate that participants communicated less

frequently with their Internet partner (5.2

times/month) than their nonInternet partner (7.2

times/month, p < 0.001) as shown in Figure 3.

Although respondents communicated more by email

with their Internet partner (p < 0.001), they commu-

nicated less using the other modalities (p < 0.001 for

face-to-face and p < 0.001 for telephone). Respon-

dents also reported feeling less close to their Internet

partner than to their nonInternet partner (p < 0.001).

Using a least squares regression analysis, we pre-

dicted psychological closeness from frequency of com-

munication for the nonInternet partner and Internet

partner, controlling for sex, age, role relation, duration

of acquaintance, and physical proximity. Most

notably, frequency of communication was a critical

predictor of psychological closeness with the nonIn-

NonInternet partner

Internet partner

Face-to-face

Phone

Total

nicating in person

(Beta s= 0.36) and by

telephone (Beta = 0.27)

were both significantly

better predictors of a strong relationship than was

communication by email (Beta = 0.15).

Comparing Internet Versus NonInternet

Social Partners

In these studies, respondents selected communica-

tion episodes and partners based on the recency of

the communication session. This procedure has the

advantage of sampling all potential conversations,

but may over-represent social relationships not

important to the respondents, but are frequent sim-

ply because the partners are nearby. Here, we com-

pare the value of using computer-mediated and

noncomputer-mediated communication to keep up

with partners with whom the respondents have a

substantial amount of communication. The data

comes from the HomeNet project, a field trial that

tracked Internet usage and communication behavior

among a sample of 93 households in Pittsburgh dur-

ing their first year or two online (see [5]).

Participants answered a series of questions about

two individuals with whom they had frequent com-

COMMUNICATIONS OF THE ACM July 2002/Vol. 45, No. 7

105

Figure 3. Frequency of

communication per month by

modality and partner type.

100%

75%

50%

25%

500

1000

1500

2000

25%

Ratio of Active Participants to Total Members

75%

100%

Membership Size

A. (N=204 groups)

B. (N=139 groups)

100%

75%

50%

25%

1.0

1.5 2.0 2.5 3.0 3.5 4.0

Average Number of Messages per Discussion Thread

Average Number of Messages Per Day

C. (N=204 groups)

D. (N=139 groups)

ternet partner

(Beta = 0.40),

but not with the

Internet partner

(Beta = -0.08).

The difference is statistically significant (p < 0.001).

The weaker association of communication with close-

ness for the Internet sample is analogous to findings

from the student sample.

Social relationships offline involve more commu-

nication than those developed online, and thus pre-

dicted psychological closeness. Given our

cross-sectional data, we cannot tell if communication

does not lead to closeness when people are communi-

cating electronically, or if people are exchanging email

with people to whom they do not feel close. In either

case, they are not getting as much social benefit from

email as they do from their other communication

activity.

these electronic or virtual communities are often

described as groups where relationships form, and

whose members provide each other with compan-

ionship, information, and social support [11].

While existing studies and stories of electronic

groups provide insight into the types of social activity

that can occur in electronic collectives, the anecdotal

nature of this research leaves open the question of

what typically happens. Are active, tightly knit elec-

tronic groups, in which people form personal rela-

tionships and develop a sense of belonging, the norm

or are the cases reported in the literature interesting

exceptions? To examine this question, we collected

data from a sample of 204 Internet listservs. The data

shows that, on average, listservs are much more like

loosely knit, voluntary organizations than the tightly

knit social communities highlighted in prior case

studies.

The sample consisted of 204 unmanaged and

unmoderated email-based listservs, drawn from a

population of approximately 70,000 listservs. An ini-

tial random sample of 1,066 was stratified by topic

type (work-related, personal, and mixed) to ensure it

included a range of topics and member populations.

Listservs were dropped from the initial sample if the

list owner declined to participate in the study (21%);

the listserv was defunct (16%); it had closed mem-

bership, generally as part of an organization, course,

or task force (15%); or it could not provide member-

ship data in an automated fashion. The final sample

consisted of lists evenly divided among those oriented

Online Social Groups

The research we described so far concentrates on

dyadic relationships between individuals in their

online and offline lives. Yet one of the prominent

features of the Internet is the presence of larger

social collectives, which researchers have called “elec-

tronic groups” or “communities.” Even before the

advent of the Web, the Internet provided an infra-

structure for online group-level social behavior,

through USENET and email-based distribution

lists. In descriptions of social life on the Internet,

106

July 2002/Vol. 45, No. 7 COMMUNICATIONS OF THE ACM

50%

Figure 4. The distribution of listservs

in membership, and participation

ratio, message volume, and discussion

length.

around professional, personal,

and academic topics. Based on

descriptions of the lists, we were

able to classify them as purely

electronic or as hybrid, combin-

ing both electronic and tradi-

tional communications, especially

conventional face-to-face meet-

ings. For example, a national list

for youth hockey was judged as

purely electronic, while the mail-

ing list for a city-specific country

dancers’ group was judged as

hybrid (see [2]).

For a 130-day period we col-

lected data on each listserv’s

membership and communica-

tion activity. During the observa-

tion period, membership was

characterized in terms of size (number of members),

growth (members entering as a percentage of initial

size), loss (members leaving as a percentage of initial

size), and net change in size (as a percentage of initial

size). Communication activity was measured in terms

of volume (number of messages per day) and interac-

tivity (length of discussion threads). In addition, mea-

sures of member participation (percentage of

members contributing messages and the concentra-

tion of message contributions among the active par-

ticipants) were created for each listserv. The table

appearing here delineates this information, and con-

trasts the purely electronic with the hybrid lists. Fig-

ures 4a–d, respectively, show the distribution of the

number of subscribers, the proportion of all sub-

scribers who posted a least one message during the

130-day observation period, the daily number of mes-

sages per listserv, and the average number of messages

per discussion thread.

Unlike traditional small groups, listervs have large,

fluctuating memberships in which a small core of

active participants generates relatively low levels of

sporadic communication, whose messages rarely

receive a response. Small groups, as described in the

social psychological research literature, have between

3 and 15 members, with relatively low turnover. By

comparison, the listservs were much larger (median of

64 members), with high churn (22% of original

members dropping out annually and double this

number joining). In contrast to highly interactive

conversation involving almost all group members

(typical of small groups), listservs exhibit little com-

munication, with a full 33% exhibiting no communi-

cation during the 130-day observation period. Of

those that did, the median listerv accrued 0.28 mes-

Variable

Measure

Mean

Overall

Median

Overall

Median

Pure

Electronic

Lists

Median

Hybrid

Lists

Different

Pure vs.

Hybrid

Size

Number of members at start of

observation period

Member entering during 130

day observation period as

percentage of initial size

Members leaving during 130

day observation period as

percentage of initial size

Net change in size as

percentage of initial size

Average number of messages

per day

Average discussion thread length.

Thread length of 1 indicates the

originating message

Percentage of members who

contributed messages during the

observation period

163

***

Growth

23%

15%

14.8

16.7

Loss

13%

8.1%

Change

Volume

1.650

0.28

0.24

0.38

Interactivity

1.580

1.33

1.61

1.50

Member

Participation

22%

15%

19%

N = 204 listservs

*** p< 0.001

33% of the sampled groups had no activity during the observation period. The reported data is from the active groups.

sages per day (or less than

0.0004 messages per sub-

scriber per day). Over

50% of members con-

tributed no messages over the 130-day observation

period, and a small number generated most of the

messages. Conversation was not interactive. On aver-

age, fewer than one message out of three received any

response.

The hybrid groups differed little from the purely

electronic groups. Though they were significantly

smaller, probably reflecting the more limited geo-

graphic area from which they could attract members,

both types of groups had similar high turnover, low

volume of messages, low level of interactivity, and

domination by a small proportion of their member-

ship. Regardless of how the hybrid groups acted when

they met face-to-face, online they acted like typical

weak-tie collectives.

In terms of membership size and change, commu-

nication volume and structure, and participation lev-

els, Internet listservs do not appear to be intimate

social groups. These findings highlight a bias in prior

research on online social activity. While the goal of

describing the existence of true social behaviors in

online environments has been well served by focusing

on highly active and interactive examples of electronic

collectives, these cases are not representative of what

typically happens. For example, interactivity is a com-

mon theme in many descriptions of online social

activity (for example, [1]). However, our results imply

that while interactivity can occur in these contexts, it

is the exception, not the rule, when it occurs.

It was not the case that all listservs in this sample

had impoverished social behavior, although this was

Population characteristics

for Internet listservs.

COMMUNICATIONS OF THE ACM July 2002/Vol. 45, No. 7

107

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: