Human communication in customer-agent-computer interaction: Face-to-face versus over telephone

Computers in Human Behavior 25 (2009) 8–20

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier .com/locate /comphumbeh

Human communication in customer-agent-computer interaction:Face-to-face versus over telephone

Anette Kira a,*, David M. Nichols b, Mark Apperley b

a AgResearch Limited, East St., Private Bag 3123, Hamilton, New Zealandb Department of Computer Science, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand

a r t i c l e i n f o

Article history:Available online 18 July 2008

Keywords:Customer-agent-computer interactionHuman communicationTelephoneFace-to-faceCommunication mode

0747-5632/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.chb.2008.05.013

* Corresponding author. Tel.: +64 7 838 5093; fax:E-mail address: [email protected] (A. K

a b s t r a c t

Customer service can be provided over various communication modes, such as face-to-face, telephone,email or websites. In this paper we examine a setting where a service, such as travel planning, is providedto a customer through a human agent, either face-to-face or via telephone. Specifically, the settingrequires three entities, a customer who has approached a business, a representative for the organisationand a computer which the representative uses to support the task. Two experiments were conducted toinvestigate how the two human entities interact over two different communication modes (face-to-faceand telephone) when there is also a computer involved in the interaction. The results showed a signifi-cantly shorter task completion time via telephone. There was also a difference in the style of communi-cation, with face-to-face having more single activities (such as talking only), while when using thetelephone there was more doubling up in activities (talking while also searching on the computer). Therewas only a small difference in subjective satisfaction. The results suggest that telephone interaction,although containing fewer communication cues (such as body language), is not necessarily an impover-ished mode. Telephone interaction is less time consuming and more task-focused.

� 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Customer service can be provided over various communicationmodes, such as face-to- face, telephone, email or websites. Histor-ically, much customer service has been provided face-to-face,however, increasingly service is provided via telephone or com-puter-based methods, such as email, web sites and instant messag-ing. In the ‘old factories’, work was concentrated on themanufacturing floor, while the ‘new factories’ concentrate on offer-ing services through customer representatives working in front of acomputer (Bowers & Martin, 2000). Many types of organisationsprovide customer service as part of their activities, including travelagencies, banks, insurance companies, supermarkets, call centres,clothes stores, car sales yards, etc. There are different kinds of ser-vices; for instance, moving a power account to a different address,designing overseas trip, or assisting customers with clothing pur-chases. The goals of the service vary with the genre of the organi-sation; electricity provider call centres focus on reducing the timerequired for answering and processing a call – whereas the travelindustry focuses on providing an enjoyable experience for thecustomer.

There is considerable debate in the literature about the pros andcons of the old and new approaches to customer services – largely

ll rights reserved.

+64 7 838 5117.ira).

centering on the efficacy of the interactive process. One process ofincreasing interest in this respect is the customer agent computerinteraction (CACI – see Fig. 1). CACI focuses on collaborative inter-actions such as purchasing a plane ticket at a travel agency. Collab-oration involves two or more people engaging in some commontask, interacting with various tools and products and communicat-ing to co-ordinate their work (Dix, Finlay, Abowd, & Beale, 1998).For example, in client service interaction, the agent needs to com-municate with the customer while also interacting with a com-puter (Bowers & Martin, 2000; Lawrence, Atwood, & Dews,1994). The customer is also trying to accomplish their goal (book-ing a flight) with the computer system, but is doing so via the orga-nisation’s representative. Hence, the agent acts as an interfacebetween the customer and organisation (Randall & Hughes, 1995).

Some research suggests that as long as people can communicateverbally, adding visual communication does not improve the inter-action (Chapanis, Ochman, Parrish, & Weeks, 1972; Gale, 1991;Ochsman & Chapanis, 1974). Additionally, it has been argued thatwhen completing a course on problem solving, Web-based interac-tion does not differ significantly from face-to-face interaction (Or-vis, Wisher, Bonk, & Olson, 2002). Chapanis et al. (1972) andOchsman and Chapanis (1974) conducted a series of studies inves-tigating communication modes, ranging from written-only to face-to-face, involving two people collaboratively solving a task, withone person being an ‘ideal computer’ and the other person the‘computer user’. An impoverished communication mode, such as

mailto:[email protected]

http://www.sciencedirect.com/science/journal/07475632

http://www.elsevier.com/locate/comphumbeh

Hc Ha C Hc Ha C

a b

Fig. 1. Comparing human (Hc)–human (Ha) interaction over two different commu-nication modes: (a) telephone (t) and (b) face-to-face (f).

A. Kira et al. / Computers in Human Behavior 25 (2009) 8–20 9

telephone or electronic communication, might not necessarily leadto a less satisfactory outcome or process of communication (Ochs-man & Chapanis, 1974; Orvis et al., 2002).

Other research, in contrast, suggests that there is a difference inoutcome, style and content between audio-only and face-to-facecommunication (Rutter, Stephenson, & Dewey, 1981; Short, Wil-liams, & Christie, 1976; Suchman, 1987). Audio-only interactionsare more task-oriented, depersonalized, less spontaneous, andend in a different outcome, and additionally audio-only interactionhas fewer interactional cues, such as body language or the activityof the other collaborator. Face-to-face communication style hasmore speech overlaps (people speaking at the same time) and ashorter pause, while audio-only has longer speech acts and fewerconversational turns, more pauses and less overlaps (Matarazzo& Sellen, 2000). Furthermore, it has also been suggested that peo-ple preferred face-to-face interactions (Fletcher & Major, 2006;Masoodian & Apperley, 1995; Masoodian, Apperley, & Frederikson,1995).

With the increasing use of CACI interactions in commerce theissue of its efficacy compared to traditional communication meth-ods is also gaining in importance. Research thus far into CACI inter-actions has been limited. In one of the few studies Steel, Jones, andApperley (2001), Steel, Jones, and Apperley (2002) and Steel (2003)investigated adding auditory feedback to the computer and cus-tomer communication channel, in a simulated utility call centresetting where the participants communicated via telephone. Thepurpose was to provide auditory information about the state ofthe computer directly to the customer. In one auditory feedbackstudy (Steel et al., 2002) there were indications that the addedauditory feedback enhanced the interaction, by improving taskcompletion time and flow of interaction. However, in the secondmore constrained experiment, there was only a very small non-sig-nificant difference in task completion time, flow of interaction andsubjective satisfaction. The conclusion from these studies was thatadding auditory feedback did not appear to have a significant effecton the interaction. Olsson (2007) observes that in order to assesswhy adding auditory feedback did not improve the interaction, fur-ther research into the communication between agent and cus-tomer is required.

The research setting described in this paper requires three enti-ties: one or more customers who have approached a business, arepresentative for the organisation (the agent) and the agent’scomputer. In addition, there is also an asymmetry between the hu-man roles, where the customer has specific details relating to thetask, and the agent has the domain expertise. The agent mediatesbetween the customer and information system. The customerwho wants a service or product has the detailed informationneeded to complete the task, but is not directly interacting withthe computer.

The purpose of these studies were to investigate if communica-tion mode, face-to-face versus telephone, has an effect on howpeople communicate in a customer service setting when there isa computer involved in the interaction. Section 3 describes the firstexperiment including method (Section 3.1), results (Section 3.2)and discussion (Section 3.3). In Section 4 a second experiment isdescribed which focused on subjective satisfaction, includingmethod (Section 4.1), results (Section 4.2) and the discussion (Sec-

tion 4.3). The combined findings from both experiments are dis-cussed in Section 5.

2. CACI interaction – a theoretical background

Those transactions where the agent acts as an interface be-tween the customer and the organisation, using a computer, area common occurrence in daily business communication. Howeverthere is a distinct absence of literature that describes and investi-gates CACI interaction. Therefore it is necessary to draw directionand corroboration from related fields, such as computer supportedcollaborative work (CSCW), customer service and human-communication.

The primary product for many commercial organisations is tooffer a service, and hence providing good customer service isessential. In the ‘old factories’ work was concentrated on the man-ufacturing floor, while the ‘new factories’ concentrate on offeringservices through customer representatives working in front of acomputer (Bowers & Martin, 2000). Millard et al. (2000) argue that70% of the customers that terminate their relationship with a busi-ness leave because of bad service. Because of the vast number ofservice interactions and their relationship to important outcomes,such as loyalty, sales and satisfaction, it is essential to understandhow to manage service encounters, even in industries that are nottraditionally defined as service industries (Bitner, Brown, & Meu-ter, 2000). Bradshaw and Brash (2001) define customer relation-ship management as

‘‘. . . a management approach that enables organisations to iden-tify, attract and increase retention of profitable customers, bymanaging relationships with them.”

The call centre is where an organisation, which does not provideface-to-face customer contact, gets an opportunity to understandthe clients’ needs, and to offer solutions and support to the organi-sation’s product and services (Tanir & Booth, 1999). It is importantto ensure that:

s The customer receives quality service.s The service is delivered promptly.s That many customers are served ‘at their convenience’.

There is a substantial amount of research into how people worktogether via computers (for example, Grudin, 1988 & Gutwin &Greenberg, 1998), and their focus has been on making electroniccollaboration easier. Research has also investigated how groupswork together and how to classify different types of collaboration(for example, Bardram (1998), Grudin (1994), Gutwin & Greenberg(1998)). The time/space matrix is the most common classificationand splits collaboration into whether the participants are workingco-located or remotely, and if they are working synchronously orasynchronously (Dix et al., 1998). For example, when two peopleare communicating face-to-face, it is co-located and synchronous.

Human-to-human communication is often complex and it canbe ambiguous and error-prone (Balint, 1996), and additionally, hu-mans only have a limited amount of cognitive resources (Walker,1994). When solving problems, communicating with other humansis only one of the activities that require cognitive resources. Com-munication has to share the available resources with other activi-ties, such as making inferences, retrieval from long term memoryand interacting with other artefacts. In certain situations, forexample business and financial administration, where correctnessand exactness is important the inclusion of a computer can assistin achieving the desired precision.

There are different formats of human communication such asface-to-face or telephone, depending on the setting and the

10 A. Kira et al. / Computers in Human Behavior 25 (2009) 8–20

artefacts in that setting (Greatbatch, Luff, Heath, & Camion, 1993;Lawrence et al., 1994; Ten Have, 1991). An example is the asym-metric communication between a patient and a doctor (Greatbatchet al., 1993; Ten Have, 1991). When a patient visits a doctor, it isthe patient’s, not the doctor’s, health that is under review (TenHave, 1991). The patient reports on their illness, answers the doc-tor’s questions and accepts (or chooses not to accept) the physi-cian’s recommendation. The doctor, on the other hand, listens,investigates and diagnoses. However, the asymmetry is notalways consistent; at times the doctor-patient communicationmay be a normal informal conversation between equals, while atother times it may be formal and asymmetrical. Participants mayalso move between the formal and informal modes during aconsultation.

One common CACI scenario can be found in telephone directoryassistance. Lawrence et al. (1994) and Lawrence et al. (1995) sug-gest that a telephone directory assistance agent is in effect a ‘sur-rogate user’ between the customer, with the question, and thecomputer containing the desired information. The agent has tobridge the ‘gulf of execution’ (Norman, 1988) between the client’slanguage and the technical terms needed for the computer. The‘surrogate user’ needs to obtain information from the customer,translate it into technical terms and search for the listing on thecomputer, and then translates the technical result back to the cus-tomer. This translation from ‘social’ to ‘technical language’ is oftennot trivial (Lawrence et al., 1994) and additionally, the agent has tovary the use of social and technical language due to the character-istics of the caller; for example, if the caller was elderly or a child,the operator reverts to a more social interaction.

Social interaction is often time consuming, but directory opera-tors are often required to complete the task as quick as possible(Lawrence, Atwood, Dews, & Turner, 1995; Muller et al., 1995).Thus to speed up task completion, ‘technical language’ is briefand factual, and is employed with little social talk (Lawrence etal., 1995). Conversational turns are short and words were oftenomitted; for example ‘‘In Brooklyn?” instead of ‘‘Is that numberin Brooklyn?”. To encourage a technical, as opposed to a socialcommunication style, the greeting can be directive and task-ori-ented, such as ‘‘What city, please?”. Furthermore, the operatorswho are better at working in parallel with the two contexts, tech-nical and social, complete the request in a shorter time. The fasteroperators are able to work with the computer and the customer atthe same time, leading to fewer interruptions. The slower opera-tors, on the other hand, work in more of a ‘response mode’, drivenby the customer, leading to more interruptions due to the agenttyping or waiting for the customer to formulate their nextutterance.

The ability of an agent to manipulate the computer system andunderstand the clients needs has led Muller et al. (1995) to suggestthat a telephone directory assistant is not just a ‘surrogate user’,but is more of a knowledge worker, in contrast with Lawrenceet al. (1994) and Lawrence et al. (1995). In addition to knowledgeof the computer system and an ability to translate between thecomputer system and customer, the operator also needs:

s Knowledge of the customer’s world, for example, knowledge ofunusually spelt locations, such as ‘‘Berkeley” being ‘‘Buerkle”.

s Changes that occurs in the customer’s world, for example, achange in the name of a government agency.

s Knowledge about information that is not stored on the com-puter database.

As well as specific knowledge, the agent also needs further abil-ities, such as, problem solving skills and interpersonal skills,including an ability to collaborate with the customer to refinethe queries.

Another social environment where CACI has been studied is be-tween the customer and agent at travel agencies (Halloran, 2002;Randall & Hughes, 1995; Rodden, Rogers, Halloran, & Taylor,2003; Scaife, Halloran, & Rogers, 2002). One of the main problemsin the interaction between the customer and agent is the asymme-try, both in the physical layout of a travel agency and how theinformation is represented for the two parties. The physical layoutat a travel agency makes it difficult for the customer to be part ofthe collaboration, even if they wanted to (Halloran, 2002; Roddenet al., 2003; Scaife et al., 2002). In a typical travel agency situation,the customer and the agent sit on either side of a desk. The agenthas access to the information system, while to the customer theagent is the interface to the computer system.

A constraint on effective collaboration can be the difference ininformation representation. The customer’s itinerary is often orga-nized chronologically, while for the agent the trip has to be workedon by product category, for instance, flights, tours or accommoda-tion (Halloran, 2002; Rodden et al., 2003; Scaife et al., 2002). Theinformation available for the customer, such as brochures or handwritten notes, is different to the information the agent uses, suchas computer-based flight booking systems. Additionally, the infor-mation on the computer is tailored for the agent and does not usu-ally account for the additional interaction between representativeand client. The interaction between the agent and customer is pre-dominantly verbal as the customer tells the agent what they wantand the agent reads out various options. However, due to the com-plex nature of travel booking, the interaction often becomes com-plicated, with the customer having to remember many differentoptions. It is common that the agent at this stage prints out a sum-mary of various options and uses the printout as a discussion point(Halloran, 2002).

People communicate via a range of methods, including spokenlanguage, non-speech sounds or body language. Co-located syn-chronous modes of communication, such as face-to-face, supportall the different types of communication cues, such as spokenand body language, while remotely located modes, for exampletelephone, only support some of the available cues, for examplespoken language only. There are both advantages and disadvan-tages with co-located synchronous (face-to-face) communication(Nardi & Whittaker, 2001); one advantage is that it supports sus-taining a social relationship which is required to enable remotelylocated collaboration. Galin, Gross, and Gosalker (2004) found thatface-to-face was substantially faster than electronic communica-tion, but there was no significant difference in outcome, such as fi-nal price and number of payment instalments. But there are alsodisadvantages with co-located communication; it requires effort,it can be disruptive and costly, since people, by definition, haveto be at the same location (Nardi & Whittaker, 2001).

Research into human–human communication appears to beinconclusive (for example; Gale, 1991; Ochsman & Chapanis,1974; Rutter et al., 1981; Short et al., 1976; Suchman, 1987),and also, in these early studies the participants did not have ac-cess to computers. In the collaborative computing studies (forexample; Bardram, 1998; Grudin, 1994; Gutwin & Greenberg,1998), both humans have access to a computer, and there is nor-mally some amount of shared information via the computerscreen. The research area discussed in this paper, the customer-agent communication channel of CACI, differs markedly fromboth human–human communication studies and collaborativecomputing studies. Unlike human–human communication stud-ies, there is a computer involved in CACI, thus providing a thirddimension to the task. CACI also contrasts with collaborativecomputing in that although there is a computer involved, onlyone of the users has access to it, there is no sharing of the infor-mation via a computer and the people are not communicating viaa computer.


The frequency of CACI transactions and its intimate relationshipwith customer service, retention and revenue for commercial andgovernmental organisations highlight the validity of and need forresearch in this field. In addition, the lack of research into CACI,in particular the customer-agent communication channel, theinconsistent findings in previous human communication studies,and the results from the auditory feedback in CACI studies all sup-port the value of research in this area.

3. Comparing human–human communication face-to-face orover telephone

An experiment was conducted as a result of the studies carriedout into the use of auditory feedback in CACI (Olsson, 2007; Steel etal., 2002; Steel et al., 2001) and the inconsistent findings from pre-vious human communication research, for example Ochsman andChapanis (1974) and Rutter et al. (1981). The experiment wasbased on Ochsman and Chapanis (1974) work and a similar designwas used. The aim was to investigate the differences in human–hu-man communication in CACI comparing face-to-face and telephonemodes where there is a computer involved in the interaction. Thehypotheses were:

� There will be a difference in task completion time between face-to-face and telephone conditions.

� There will be a difference in style and content conditionsbetween the face-to-face and telephone condition.

� There will be a difference in subjective satisfaction betweenface-to-face and telephone condition.

3.1. Method

The experiment was a between-subject study, where partici-pants performed one task either face-to-face or via telephone.

3.1.1. ParticipantsIn each session there were two participants, one person was

randomly assigned the role of agent and the other the role of cus-tomer. There were 19 participant pairs in total, 9 in the face-to-facecondition, and 10 in the telephone condition. Due to technicalrecording problems there was an imbalance in the number of par-ticipants between the experimental conditions. There were aneven distribution of male (20) and female (18) participants of vary-ing cultures including New Zealander, Asian and European. Theages varied from 19 to 49 and educational lever varied from no ter-tiary education to Bachelor or higher.

3.1.2. TasksOchsman and Chapanis (1974) used a number of criteria for

selecting their tasks, because and a similar task selection processwas used for this study. Based on Ochsman and Chapanis’ work,the following criteria were used for this study:

� The problem was practical and real.� The problem did not require any special skill or knowledge for

the participating ‘‘actors”.� Both participants needed to collaborate and communicate to be

able to solve the task.

Additional to Ochsman and Chapanis’ criteria, for this study itwas important that the tasks were not too short to ensure that en-ough communication data could be collected. The task selectedwas a flight booking task. The customer had a range of criteriafor the flight booking and the agent had to find a flight that best

fulfilled these criteria. These criteria included the departure dateand time, and the destination.

In a real situation, a customer decides when an option has beenfound that most satisfies their needs. Hence, to give the study asmuch external validity as possible, the customer was given theresponsibility to decide when the task was completed.

The agent and customer were given separate instructions. Theinstructions included all the information needed to perform thetask, but also restrictions. It was strongly emphasized that the par-ticipants were only allowed to communicate verbally and theywere not allowed to share any notes, computer screen or other re-sources. The instructions for the agent included information aboutthe websites, and also restrictions on what could be done. Forexample, ‘‘the customer has to use the same airline in bothdirections”.

3.1.3. Data collectionTo address the three hypotheses a range of data was collected:

� Difference in task completion time.s Task completion time.

� Difference in style and content conditions.s Number of correctly selected flight details.s Activity sampling.s Duration of utterances of time in conversation.

� Difference in subjective satisfaction.s Subjective satisfaction questionnaire data.

The selection of data collection was based on previous humancommunication research by Ochsman and Chapanis (1974),namely, task completion time and activity sampling. However,since other research (for example, Rutter et al., 1981) suggestedthat face-to-face communication resulted in a different outcomeand a different communication style, selected flight details (out-come) and utterances data was also collected. In addition, becauseit has also been argued that people prefer face-to-face communica-tion (Fletcher & Major, 2006; Masoodian et al., 1995) it wasdecided to also collect subjective satisfaction.

3.1.3.1. Task completion time. The task completion timing startedwhen the participants initiated the conversation and finishedwhen the customer stated that the task was completed. In the tele-phone condition, the timing started as the customer established atelephone connection with the agent, and ended as the customerhung up the phone.

3.1.3.2. Selected flight details. At the end of the session the customerwas asked to state the outcome of the interaction. The task hadbeen set up such that there was only one correct option for eachquestion. The number of correct options was counted, with fivemeaning that they were all correct and zero meaning none of thedetails were correct.

3.1.3.3. Activity sampling. An activity sampling method was used tocollect qualitative information about the interaction. The activitythe participants were engaged in was recorded every fifth second.Not all categories used by Chapanis et al. (1972) were relevant tothis experiment, for example ‘handling parts’ as there were nophysical parts that the participants needed to handle to completethe tasks. Based on Chapanis et al.’s work, 12 categories were usedin this experiment to classify the activities (see Table 1).


3.1.3.4. Speech utterances pattern. Data about speech utterances,such as duration, place in communication, which person madethe utterance, was collected, based on Rutter and Stephenson’s(1977) experiment. Each time a participant started an utteranceand when the participant stopped the utterance was recorded.An utterance was defined as a continuous string of speech, irre-spective of whether this was a complete sentence or part sentence.There could be an overlap in utterances, when the participantstalked at the same time. Examples of utterance included:

� ‘I am trying to find a flight with Qantas now’ or,� ‘I am just trying’.The speech utterance data contained two

subsets:� Duration of utterances; it was recorded when the utterance

started and finished. The length of that utterance was calculated.For example:

� The time in the conversation where the utterance took place.The middle point of the utterance was calculated using the startand end of the utterance and was rounded to the nearest wholesecond.

3.1.3.5. Subjective satisfaction. Subjective satisfaction was collectedvia a post-task questionnaire. The participants were asked to fillout a questionnaire after the completion of the task. The question-naire included questions such as the level of frustration, satisfac-tion, understanding etc. and was similar to the one used byMasoodian et al. (1995). All questions were on a 7 point Likertscale, with 1 being the least favourable rating and 7 being the mostfavourable rating.

3.1.4. Materials3.1.4.1. Location. The experiment was held in a usability laboratoryusing two rooms for the participants. For the face-to-face conditionboth participants were located in a single room, with the customersitting on the opposite side of a table and so not able to see thecomputer screen. For the telephone condition, the agent waslocated in the first room and the customer in a second room.

3.1.4.2. Hardware. For the telephone condition a hands-free tele-phone was used by the agent and a standard telephone by the cus-tomer. For the face-to-face condition there was no addedcommunication hardware. For both conditions, the agent used astandard desktop PC.

3.1.4.3. Software. The agent used the online booking sections of thewebsites of Air New Zealand (www.airnz.co.nz) and Qantas(www.qantas.co.nz), see Fig. 2.

All sessions were recorded the video recordings contained thecomputer screen and the faces of the customer and the agent.The associated audio recording included both the customer andthe agent.

Table 1Activity classifications used to classify the participants’ activities

Single activities Dual activities

Talk only Talking and searchingTalk pause Talking and making notesListening only Listening and sendingSearch only Listening and searchingMaking notes Listening and making notesWaiting Searching and making notesOther

3.1.5. ProcedureThe procedure for each session was:

� Participant who was assigned the job of agent was greeted byresearcher.

� The agent was given a short training session on the websites, aswell as five minutes to familiarize themselves with the sites.

� The other participant, the customer, arrives.� The procedure was simultaneously explained to both partici-

pants, followed the ethical consent procedure, which includedinformation that the study was about communication.

� The participants were given written instructions.� The participants performed the task.� The participants filled out a questionnaire.

For both conditions the participants were only allowed to com-municate verbally. They were not allowed to share any notes, writ-ten or typed, nor were they allowed to share the computer screen.

3.2. Results

This section covers each of the collected data sets.

3.2.1. Task completion timeThe mean task completion time for the telephone condition was

1148 s, while it was 1937 s for the face-to-face condition (see Table2). However, the standard deviation was larger for the telephonecondition (655 s) than the face-to-face condition (457 s).

The task completion time data was continuous but not normallydistributed, consequently a Wilcoxon Rank-Sum test was used forcalculating statistical significance (Bhattacharyya & Johnson,1977). The sum of the ranks was 120 for the face-to-face conditionand 70 for the telephone condition. The probability of obtaining arank sum of 119 or greater on a sample of 19 is less than 0.009.Hence the difference between the conditions was statistically sig-nificant (Wilcoxon Rank-Sum, N = 19, p-value = 0.018).

Most of the face-to-face pairs had longer task completion timesthan the pairs in the telephone condition. Seven of the 10 partici-pant pairs in the telephone condition had shorter task completiontimes than any participant in the face-to-face condition (see Fig. 3).

3.2.2. OutcomeDescriptive statistics were calculated for the outcome, and there

was only a small difference between the two conditions. Overall,the face-to-face condition had a slightly higher mean number ofcorrectly selected criteria. The difference was very small thoughwith the mean for face-to-face 4.1 and for telephone 3.7.

3.2.3. Activity samplingAs described in the method section, the activity sampling was

taken every five seconds of the entire conversation. Descriptivestatistics were calculated for activity sampling (see Table 3). Be-cause the task completion time was longer for the face-to-face con-dition, the data was normalized. Overall the face-to-face conditionhad more occurrences of single activities, while the telephone con-dition had more occurrences of dual activities (see Table 4).

For all the single activity categories, the face-to-face conditionhad a higher count (see Fig. 4). In contrast, for all dual activities,the telephone condition had a higher count. Some categories con-tributed more to the overall difference.

As the data was categorical a chi-squared test was used to cal-culate the statistical significance of the activity sampling (Bluman,2001). The result for the activity sampling was statistically signif-icant, at 0.01 level (chi-squared, N = 19, p = < 0.001). A chi-squaredtest also helps to indicate which categories contributed to the p-va-lue. The main contributing categories were ‘Listen and Search’, ‘Just

http://www.airnz.co.nz

http://www.qantas.co.nz

Fig. 2. Screen shot of the search page for the Qantas website during the time of the study.

Table 2Descriptive statistics of task completion time

Face-to-face N = 9 Telephone N = 10

Mean 1937 1148Standard deviation 457 655Median 1770 900Maximum 2575 2250Minimum 1455 525

Table 3Descriptive statistics of single and combined activities for both conditions

Face to face N = 9 Telephone N = 10

Single Combined Single Combined

Mean 34.25 24.59 15.72 19.93Standard deviation 51.8 36.31 38.43 34.51Median 10 6 1.5 3

0

500

1000

1500

2000

2500

3000

0 2 4 6 8 10 12

Participant Number

Face-to-face

Telephone

Len

gth

of

task

co

mp

leti

on

tim

e (s

eco

nd

s)

Fig. 3. The total task completion time for each participant pair in both conditions.


Listen’ and ‘Just Talk’, where both conditions contributed. The tele-phone condition had a lower than expected (based on the chi-squared calculation) observed count for all the single activities,while the face-to-face condition had a higher than expected ob-served count for all single activities. In contrast, for the dual activ-ities, the face-to-face condition had a lower than expected count on

four of five categories, while the telephone condition had a higherthan expected count on four of five categories.

3.2.4. Speech utterances over time in conversationData pertaining to utterances over time in the conversation

were also collected.

Table 4Normalized data (over task completion time) descriptive statistics of single andcombined activities for both conditions

Face to face N = 9 Telephone N = 10

Single Combined Single Combined

Mean 0.08 0.06 0.06 0.09Standard deviation 0.12 0.09 0.11 0.13Median 0.02 0.01 0.00 0.01

Table 5Descriptive statistics of the number of utterances normalized over task completiontime

Face-to-face N = 6 Telephone N = 7

Mean 0.16 0.16Standard Deviation 0.02 0.03Median 0.15 0.15Maximum 0.18 0.23Minimum 0.13 0.12

0

1

2

3

4

5

6

0 50 100 150 200 250Number of utterances

Face-to-faceTelephone

Mea

n le

ngth

of e

ach

utte

ranc

e (s

econ

ds)

Fig. 5. Average length of utterances against the number of utterances for eachcondition.


3.2.4.1. Duration of utterances. Descriptive statistics about the dura-tion of each individual utterance was calculated. The mean dura-tion of utterances was longer for the telephone condition,3.30 seconds versus 2.81 for the face-to-face condition. Therewas large standard deviation for both conditions. For both condi-tions the majority of utterances were of short duration, less thanthree seconds. Additionally there was only a small difference be-tween the roles.

3.2.4.2. Count of utterances. Data was collected regarding the num-ber of utterances. The data was normalized by dividing the numberof utterances for each participant by the task completion time forthat participant (see Table 5). There was no difference in normal-ized number of utterances between the conditions.

3.2.4.3. Comparing duration and number of utterances. A comparisonwas made of the duration of utterances against the number ofutterances. For each participant the duration of utterances andthe average number of utterances was calculated and plotted(see Fig. 5). The telephone condition had more data points in theupper left corner, that is, utterances of longer duration and lowerfrequency. In contrast, the face-to-face condition had more datapoints in the lower right corner, that is, higher number of utter-ances but utterances of shorter duration.

3.2.5. Subjective satisfactionThere was only a small difference in the questionnaire response

between the conditions. The overall mean rating for the face-to-face condition was 4.16 and for the telephone condition it was4.67 (7 was the most favourable rating, see Section 3.1.3). Standarddeviation, median, maximum and minimum values were all verysimilar between the conditions.

There were small differences between the conditions for indi-vidual questions (see Fig. 6). For most of the questions, the tele-

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Just

talk

Talk

paus

e

Just

listen

Waiting

Search

Make n

otes

Ot

Talk

Activity

Nor

mal

ized

cou

nt o

f eac

h ac

tivity

Fig. 4. Normalized count of each

phone condition had a more favourable rating. In particular, theratings for satisfaction of conversation, level of frustration, taskcomplication and task completion were more favourable for thetelephone condition. For understanding the conversation and for‘the expressiveness of the conversation’ the face-to-face conditionhad a slightly more favourable rating.

Statistical significance was calculated using a Mann-Whitney Utest, as the data was continuous, normal distribution could not beassumed, and it was collected from two independent samples(Bhattacharyya & Johnson, 1977). There was only one significantdifference between the conditions, which was for question 3(‘‘Was the length of completing the task as expected?, p = 0.05).None of the other questions showed a significant difference.

her

and se

arch

Talk an

d make n

otes

Listen a

nd se

arch

Listen a

nd m

ake n

otes

Listen

and t

alk

Search

and m

ake no

te

Face to faceTelephone

activity for both conditions.

0

1

2

3

4

5

6

7

The

conv

ersa

tion

was

satis

fact

ory

Did

you

feel

frus

trate

d

Was

the

leng

th o

f com

plet

ing

the

task

as

expe

cted

Was

the

prob

lem

s so

lved

effic

ient

ly

Was

the

prob

lem

com

plic

ated

Did

you

com

plet

e th

e ta

sk

Are

you

conf

iden

t the

task

was

com

plet

ed a

s yo

u ex

pect

ed

Did

you

und

erst

and

the

conv

ersa

tion

The

conv

ersa

tion

was

con

cise

The

conv

ersa

tion

was

expr

essi

ve

Question

Face-to-face

Telephone

Mea

n ra

ting

for e

ach

ques

tion

Fig. 6. Mean rating by each condition for all questions.


3.3. Discussion

There were three hypotheses for this experiment. The hypothe-sis that there would be a difference in task completion time be-tween face-to-face and audio-only condition was supported. Onaverage the face-to-face condition had a significantly longer com-pletion time than the telephone condition. Moreover, most peoplein the telephone condition had shorter task completion times thanany of the participants in the face-to-face condition.

The hypothesis that there would be a difference in style andcontent conditions between the face-to-face and audio-only condi-tion hypothesis was supported. There was a statistically significantresult in the activity sampling, in contrast with Chapanis et al.’s(1972) work. The participants in the face-to-face condition spentmore time in single activities, such as ‘just talk’ and ‘just listen’,while in the telephone condition they spent more time in dualactivities, such as ‘talk and searching’ and ‘listen and search’. Theduration of utterances was slightly shorter for the face-to-face con-dition, which was consistent with Rutter and Stephenson’s (1977)findings. Additionally, the number of utterances was slightly high-er for the telephone condition, which was also consistent with Rut-ter and Stephenson’s (1977) results. The face-to-face condition hada high count of utterances that were short in duration, while thetelephone condition contained fewer number utterances but theywere longer in duration, which is consistent with Matarazzo andSellen’s (2000) work.

However, the hypothesis that there would be a difference insubjective satisfaction between face-to-face and audio-only condi-tion was not supported. There was only a small difference in sub-jective satisfaction between the face-to-face and telephoneconditions. For 7 of the 10 questions, the subjective rating wasmore favorable for the telephone condition. Further, the ratingfor the question relating to satisfaction of conversation was slightlymore favorable for the telephone condition, which contrasts withprevious research (Masoodian & Apperley, 1995).

Through this experiment a better understanding of human–hu-man communication in CACI has been achieved, and it appears thatthe mode has an effect on communication. The participants in theface-to-face condition took longer to complete the tasks and theyspent more time in single activities, such as just talking. In con-trast, there was no difference in the subjective satisfaction be-tween the two conditions. The next section discusses a secondexperiment into human–human communication. Because this

study described was a between-subject study and there was anunexpected finding regarding subjective satisfaction, it wasdecided to conduct a within-subject study that focused on subjec-tive satisfaction, including asking participants to compare the twoconditions.

4. Subjective satisfaction in human–human communication:face-to-face or via telephone

This section reports on a study into the human–human commu-nication channel in CACI (see Fig. 1). In the human–human com-munication between-subject experiment discussed in Section 3,there was only a small difference in subjective satisfaction be-tween the face-to-face and telephone conditions. This findingwas not consistent with previous research, which found that par-ticipants preferred face-to-face interaction (Masoodian & Apper-ley’s, 1995; Masoodian et al., 1995; Fletcher & Major, 2006). Tofurther investigate if there is a difference in subjective satisfactionin a customer service CACI a within-subject experiment wasconducted.

The hypothesis was that there would be a higher subjective sat-isfaction in the face-to-face than the telephone communicationmode using a within-subject design.

4.1. Method

This was a within-subject design, where all participants per-formed two tasks, one task face-to-face and one over thetelephone.

4.1.1. ParticipantsAs with the human–human communication study in Section 3,

there were two people in each session, one agent and one cus-tomer. There were 16 participants, eight females and eight males,their cultures included European, New Zealander, Asian and Indian,with ages varying from 18 to 45. Participants were recruited viaposters and email lists.

4.1.2. Experimental designFour of the participant pairs performed the face-to-face task

first, while four pairs did the telephone task first.To limit any learning-effects there were two different tasks. To

reduce any risk of bias due to task order, the order of the tasks was

Table 6The order of the conditions and tasks for four participant pairs

First task Condition Second task Condition

Participant pair 1 A Face-to-face B TelephoneParticipant pair 2 A Telephone B Face-to-faceParticipant pair 3 B Face-to-face A TelephoneParticipant pair 4 B Telephone A Face-to-face


altered, 50% of the participants did task A first, while 50% did task Bfirst. The conditions were also alternated (see Table 6).

4.1.3. TasksThe tasks were based on the same criteria as the tasks in the

previous experiment:

� The problem was practical and real.� The problem did not require any special skill or knowledge.� Both participants needed to collaborate and communicate to be

able to solve the task.

However, because this study was a within-subject study twotasks were needed. Care was taken to ensure that the tasks werebased on the same criteria as the experiment discussed in Section3 and were comparable, in expected task completion time andcomplexity.

The agent and customer were given separate instructions for thetasks. The instructions included all the information needed to per-form the task, but also restrictions. As with the human–human com-munication (see Section 3), it was emphasized that the participantswere only allowed to communicate verbally and that they were notallowed to share any notes, or turn around the computer screen.

4.1.4. Data collectionUnlike human–human communication study reported on in

Section 3, only subjective satisfaction through questionnaireswas collected for this experiment, where the participants wereasked to fill out a post task questionnaire after each task. The ques-tionnaire after each task was identical to the one used in the pre-vious study, which was similar to the one used by Masoodianand Apperley (1995).

At the end of both sessions, the participants were also asked tocomplete a comparative questionnaire. There were 11 questionsaddressing issues such as frustration level, satisfaction and effi-ciency (for example see 79). The participants were asked to ratetheir relative preference for each communication mode. Two datasets were collected from the comparative questionnaire:

� Which condition had the highest rating?� A number between 1 and 7 was assigned to the two conditions

depending on the location on a rating line. In Fig. 7, for the ‘rel-ative preference’-question, the participant’s response would

Fig. 7. Example comparative

have been recorded a five for the ‘t’-rating and a 4.5 for the ‘f’-rating. Numbers were superimposed, with the aid of a transpar-ency, on the questionnaire to obtain the number.

The distance between the two conditions on the ranking linewas calculated, in the example shown in Fig. 7, for the last questionthe difference would be two (approximately six minus four). Tosimplify discussion the distance between where the participantshad placed the ‘t’ and ‘f’ will be referred to as the rating distance.

4.1.5. MaterialsAs with human–human communication study discussed in Sec-

tion 3, the experiment was conducted in a usability laboratory,using two rooms and the same hardware.

4.1.6. SoftwareThe agent used the Air Ne Zealand and Qantas online booking

sites. There had been some cosmetic changes to the website fromthe first experiment, however, the task flow of the sites were stillthe same. All sessions were recorded, and contained both videoand audio of the customer and the agent.

4.1.7. ProcedureAll participant pairs were greeted together. Hence, even when

the participants performed the telephone task first, they had metprior to performing the task. This decision was made as otherwisethere would be a larger difference in the established rapport be-tween the participants depending on whether they performedthe telephone task or face-to-face task first.

The procedure for each session was:

� Both participants were greeted.� The procedure was simultaneously explained to both partici-

pants, followed the ethical consent procedure, which includedinformation that the study was the topic of communication.

� Participants performed a task in one condition, followed bycompletion of a post-task questionnaire.

� Participants performed a task in second condition followed bycompletion of a post-task questionnaire.

� Participants completed comparative questionnaire.

4.2. Results

Subjective satisfaction data was collected through question-naires. There were two types of questionnaires: post-task ques-tionnaires after each task, and a comparative questionnaire. Theremainder of this section describes these sets of data.

4.2.1. Post-task questionnairesDescriptive statistics for the post-task questionnaire ratings

were calculated. There was only a small difference between the

questionnaire question.

Table 7Mean rating for each question in the post-task questionnaire for both conditions

Face-to-face Telephone Difference

Satisfaction 5.67 5.63 0.04Frustration 5.38 5.22 0.16Expected length 3.92 3.91 0.01Efficiency 4.39 4.41 0.02Complicated 5.56 5.40 0.16Completion 6.83 6.54 0.29Confidence in completion 6.21 6.07 0.13Understood 6.64 6.38 0.26Conciseness 5.70 5.32 0.38Expressiveness 4.73 4.42 0.31

Table 9Descriptive statistics of the rating distance for the comparative questionnaire

Face-to-face Telephone

Mean 5.28 4.70Standard deviation 0.97 1.17Median 5.66 5.20


conditions with the mean and median rating being slightly morefavorable for the face-to-face condition. The variance was similarfor both conditions.

The mean rating for each question was also calculated (Table 7).For most questions the rating was slightly more favorable for theface-to-face condition. However, the difference was only small.

The mean rating for all questions and each participant and con-dition was calculated. Five participants rated the telephone condi-tion higher and nine participants rating the face-to-face conditionhigher. However, for eight of the participants there was only asmall difference in the mean overall rating.

4.2.2. Comparative questionnaireThe final questionnaire asked the participants to compare the

two conditions. The higher-rated condition for each question andparticipant was collected and counted. For example, if the partici-pant had rated as shown in the ‘relative preference’-question in Fig.7 one ‘t’ would have been counted. More participants rated theface-to-face condition as more favorable. The face-to-face condi-tion was rated higher 102 times as opposed to 68 for the telephonecondition.

There appeared to be large individual differences. Five partici-pants had rated face-to-face higher over 75% of the time, whilethree participants had rated the telephone condition higher 75%of the time. Only two participants had rated the conditions higherequally often (customer 5 and 7).

The number of times each condition was rated higher for eachquestion was also calculated (see Table 8). For most of the ques-tions the face-to-face condition had greater number of higher rat-ings than the telephone and the only question where telephonehad a greater number of higher ratings was for ‘‘efficiency”.

In terms of the rating distance, there was only a small differencebetween the conditions, with the face-to-face condition having aslightly more favourable rating (see Table 9).

For most of the individual questions, the rating distance wasslightly more favourable for the face-to-face condition (see Fig.8). The only question that had a higher mean rating distance forthe telephone condition was the completion level.

Table 8Number of higher rates for each individual question

Question Face-to-face Telephone

Preference 11 5Satisfaction 10 5Frustration 10 6Efficiency 7 9Effectiveness 8 8Ease of solving 8 7Completion 9 6Being at ease 9 6Understanding 10 5Engagement 9 6

4.3. Discussion

The hypothesis for this experiment was that there would be ahigher subjective satisfaction for the face-to-face than the tele-phone mode. For the post-task questionnaires there was only asmall difference between the modes, with the face-to-face havinga slightly higher mean rating on the post-task questionnaire. Forall questions the difference in the mean rating between the condi-tions was less than 0.4, and there was only a small difference foreach individual participant, with 11 having less than 1.0 point dif-ference in rating between the conditions. For the comparativequestionnaire the participants had rated the face-to-face conditionbetter more often (6.38 versus 4.25). However, there was only asmall difference in the mean rating distance between the two con-ditions, 5.28 for face-to-face versus 4.70 for the telephone. Thesefindings are consistent with the first human–human communica-tion (see Section 3), but inconsistent with other research (forexample, Masoodian & Apperley, 1995).

When the participants were forced to compare the two condi-tions, there was a difference in the rating between the conditions.But the difference appeared to be due to individual preference,where the participants would either rate face-to-face or telephonehigher for most or all of the questions asked. Five participants ratedthe face-to-face higher for over 75% of the questions and three par-ticipants rated the telephone condition higher for over 75% of thequestions. In contrast, the analysis into the mean rating distanceshowed only a small difference. For ten participants there was lessthan one difference in the rating distance. When investigating boththe highest rated condition and the rating distancebetween them,there were also large individual differences, where some partici-pants had rated one condition higher most of the time, but therewas only a small difference in the rating distance. For example cus-tomer 1 had rated telephone higher for all questions, but there wasonly a 0.9 difference in the mean rating distance. In contrast, otherparticipants had a large difference between the rating distancemeans. For example, agent six had rated face-to-face higher forall questions and the difference between the rating distances ofthe two conditions was 3.3.

The aim of this study was to clarify the finding from the hu-man–human communication described in Section 3. The resultsfrom this study agree with those findings confirming only a smalldifference in subjective satisfaction with the face-to-face conditionbeing slightly more favorable than the telephone condition. Thedifference in the mean rating of the post-task questionnaires be-tween the conditions was smaller for the study described in thissection. In the previous study it was found that telephone commu-nication is more task-focused and efficient. It is possible that peo-ple are equally satisfied with telephone as face-to-face, eventhough telephone interaction is more task-focused, with lesssmall-talk and a less interactive style of communication.

5. Overall discussion

The findings from these experiments have added to the under-standing of human–human communication in CACI. Providinggood customer service is crucial to many organisations and it isbecoming more common that this service is provided through call

0

1

2

3

4

5

6

Pref

eren

ce -

task

Pref

eren

ce -

Gen

eral

Satis

fact

ion

Free

of

frust

ratio

n

Effic

ienc

y

Effe

ctiv

enes

s

Ease

of

solv

ing

Com

plet

ion

leve

l

Bein

g at

eas

e

Und

erst

andi

ng

Enga

gem

ent

Question

Face-to-faceTelephon e

Ave

rage

ratin

g fo

r eac

h qu

estio

n

Fig. 8. Mean rating distance for each individual question.


centres. The results from these studies favour call centres and rep-licates the result that Rutter et al. (1981) found, that face-to-facecommunication is more time-consuming and less efficient, butwith no significant difference in subjective satisfaction or outcome.It is possible that the higher number of single activities for theface-to-face condition, while the higher number of dual activitiesfor telephone would contribute to the shorter task completion timefor the telephone condition. In the face-to-face condition, the par-ticipants would stop working on the current task and look at theother participant while he or she was talking. In the telephone con-dition, in contrast, people kept working on the task while also talk-ing or listening to the other participant. It is also possible thedifference in style of communication contributed to the short taskcompletion time. The face-to-face condition appeared to be moreinteractive, with participants predominantly communicatingthrough a high count of utterances that were of short duration.In contrast, in the telephone condition the participants were lesslikely to interrupt the other participant which led to the longer,but fewer, utterances, which is consistent with Matarazzo and Sel-len’s (2000) findings. In the face-to-face condition the participantsasked for more clarifications, sought confirmation that the otherperson had understood and also offered more confirmation ofunderstanding.

In the experiment discussed in Section 3 the subjective ratingsfor were slightly more favourable for the telephone condition, butthe difference was only small. In the experiment reported on inSection 4 the rating was slightly more favourable for the face-to-face condition, but again the difference was very small. Based onprevious research (for example, Masoodian & Apperley, 1995) itwas expected that the face-to-face condition would have a morefavourable rating. There are several potential reasons why thehypothesis was not supported. The people in these studies didnot have a vested interest in the outcome and this could have af-fected their lack of preference of communication modes. As theparticipants did not have a personal interest in the outcome, theinability to establish the accuracy of the outcome might not affectsubjective satisfaction. Because these studies were conducted in asimulated setting, it is also possible that participants settled for anoption that was only average as opposed to an option that fully sat-isfied all the requirements, and this could have simplified the taskand hence had an effect on subjective satisfaction. Additionally, thenature and complexity of the task was solved equally sufficientlyby either face-to-face or via the telephone. Further, it is possiblethat the restrictions in information-sharing reduced some of theadvantages of being face-to-face. With the growth of call centres

and other business transactions taking part via telephone, peoplemight be more used to conducting their business via telephone.This change in the characteristics of participants could also haveled to the lack of difference between the conditions.

There were some threats to generalisability in these studies be-cause some of the constraints, the inexperience of the agent andthe restriction on information sharing (see Section 3.1.5), werenot realistic. In most CACI settings one person is an experienceduser of the computer system and also trained in communicatingwith a customer. Further, a call centre agent has often had cus-tomer service training, including training on what to say and alsohow to interact with the customer. In these experiments neitherof the participants were experts. The agents in these experimentsdid not have any customer service training and only very shorttraining on the websites. However, to improve control in theexperiments, it was decided that the short training was an accept-able compromise. It was not possible to find enough trained travelagents to participate in the experiments. If it was the same personin the role of agent for all sessions there would have been otherthreats, such as learnability and individual differences. Further,the fact that neither participant was trained might have mitigatedany differences between the conditions.

The experimental constraints could also have had an effect onthe results. As mentioned before the participants were only al-lowed to communicate verbally and were not allowed to shareany written or electronic information. This constraint was addedto control the experiments and make sure that the two conditionswere comparable. However, in a real situation the participantswould be able to share written information, as well as sometimesbeing able to share the computer screen. In the face-to-face condi-tion, being able to share written information might have led to areduction in time, due to the customer looking at a paper itinerarywhile the agent searched for additional options.

Findings from research into human–human communication isinconsistent, where some researchers (for example Chapanis etal., 1972) argue that there is no significant difference betweenco-located and audio-only communication, while other (for exam-ple Rutter et al., 1981) suggest that there is a difference in style andoutcome between face-to-face and audio-only communication.The research in this paper has shown that for simpler tasks or ifthe customer has a lower vested interest, a less-rich communica-tion mode, telephone, is more efficient and task-oriented thanface-to-face. It also does not have a lower subjective satisfactionfor the customer or agent, and does not necessarily lead to a neg-ative outcome or process of arriving at the outcome. People engage


in multiple activities over the telephone, for example, readinginformation while also talking. In contrast, when communicatingface-to-face people often suspend the task they are working onto focus their entire attention on the other person. But the findingsalso suggest that in a more emotional or complex task there mightbe a difference in style and subjective satisfaction due to the com-munication mode.

The findings from these experiments suggest a number of areasfor future work. The current experiments were a controlled labora-tory study, and there was no reward for the participants dependingon the outcome of the tasks. Possible future studies include inves-tigating whether the mode of communication has an effect on sub-jective satisfaction in either a real situation or in a situation wherethe participants had a vested interest.

It is also possible that the type of setting affected the result andthe findings might have been different in a more emotional task,such as visiting a doctor, or a more complicated task, such as book-ing an extensive trip around the world. Future work includes inves-tigating if the nature of the task has an impact on the mode ofcommunication.

6. Conclusion

The results of investigations into customer–agent interactionshowed that telephone communication is more efficient and lesstime-consuming than face-to-face communication and the taskcompletion time was significantly shorter for the telephone com-munication. People engage in multiple activities over the tele-phone, for example, searching on the computer while alsotalking. In contrast, when communicating face-to-face people oftendiscontinue the task they are working on to focus their entireattention on the other person. Via the telephone one person wouldtalk for longer, without interruptions from the other person,whereas while interacting face-to-face people would talk for short-er periods of time with more turn-taking from the other person.These findings are consistent with Rutter et al. (1981) who arguethat face-to-face interaction is more time-consuming and less effi-cient than telephone interaction. The results also showed thatthere was only a small difference in the subjective preferences,such as satisfaction, frustration and confidence in completion.Additionally, interaction via telephone does not necessarily leadto a negative outcome or process of arriving at the outcome. Thefindings from these studies suggest that call centres are efficientin providing customer service, without any reduction in customersatisfaction.

References

Balint, L. (1996). Computer-mediated interpersonal communication: The HCHIapproach. In D. L. Day & D. K. Kovacs (Eds.), Computers, communication & mentalmodels (pp. 28–35). Bristol, PA, USA: Taylor & Francis, Inc.

Bardram, J. (1998). Designing for the dynamics of cooperative work activities. InProceedings of the 1998 ACM conference on computer supported cooperative work(CSCW ‘98) (pp. 89–98). New York, NY, USA: ACM Press.

Bhattacharyya, G. K., & Johnson, R. A. (1977). Statistical concepts and methods. NewYork, USA: John Wiley and Sons.

Bitner, M. J., Brown, S. W., & Meuter, M. L. (2000). Technology infusing in serviceencounters. Journal of the Academy of Marketing Science, 28(1), 138–149.

Bluman, A. G. (2001). Elementary statistics: A step by step approach (4th ed.). NewYork, USA: McGraw-Hill Companies Inc.

Bowers, J., & Martin, D. (2000). Machinery in the new factories: interaction andtechnology in a banks telephone call centre. In Proceedings of the 1998 ACMconference on computer supported cooperative work (CSCW ‘98) (pp. 49–58). NewYork, NY, USA: ACM Press.

Bradshaw, D., & Brash, C. (2001). Managing customer relationships in the e-businessworld: How to personalise computer relationships for increased profitability.International Journal of Retail and Distribution Management, 29(29), 520–529.

Chapanis, A., Ochman, R. B., Parrish, R. N., & Weeks, G. D. (1972). Studies ininteractive communication: I. The effect of four communication modes on the

behavior of teams during cooperative problem solving. Human Factors, 14(6),487–509.

Dix, A., Finlay, J., Abowd, G., & Beale, R. (1998). Human–computer interaction (2nded.). Glasgow, UK: Pearson Education Ltd.

Fletcher, T. D., & Major, D. A. (2006). The effect of communication modality onperformance and Self-rating of teamwork components. Journal of Computer-Mediated Communication, 11(2), 557–576.

Gale, S. (1991). Adding audio and video to an office environment. Studies in computersupported cooperative work: Theory practices and design. Amsterdam, TheNetherlands: North-Holland Publishing Co.

Galin, A., Gross, M., & Gosalker, G. (2004). E-negotiation versus face-to-facenegotiation what has changed – if anything? Computers in Human Behavior,23(1), 787–797.

Greatbatch, D., Luff Heath, C., & Camion, P. (1993). Interpersonal communicationand human computer interaction, an examination of the use of computers inmedical consultations. Interacting With Computers, 5(2), 193–216.

Grudin, J. (1988). Why CSCW applications fail: Problems in the design andevaluation of organizational interfaces. In Proceedings of the 1988 ACMconference on computer-supported cooperative work (CSCW’88) (pp. 85-93).New York, NY, USA: ACM Press.

Grudin, J. (1994). Computer-supporter cooperative work: history and focus.Computer, 27(5), 19–26.

Gutwin, C., & Greenberg, S. (1998). Design for individual, design for groups:Tradeoffs between power and workspace awareness. In Proceedings of the 1998ACM conference on computer supported cooperative work (CSCW’98) (pp. 207–216). New York, NY, USA: ACM Press.

Halloran, J. (2002). Putting it all together: Information visualizations, displayarrangements, and sales transactions. In New Review of Information Networking,8, 3–31.

Lawrence, D., Atwood, M.E., & Dews, S. (1994). Surrogate users: Mediating betweensocial and technical interaction. In Proceedings of the SIGCHI conference onhuman factors in computing systems: Celebrating interdependence (CHI’94) (pp.399–404). New York, NY, USA: ACM Press.

Lawrence, D., Atwood, M. E., Dews, S., & Turner, T. (1995). Social interaction in theuse and design of a workstation: Two contexts of interaction. In P. Thomas (Ed.),The social and interactional dimensions of human computer interfaces. Cambridge,UK: Cambridge University Press.

Masoodian, M., & Apperley, M. (1995). User perceptions of human-to-humancommunication modes in CSCW environments. In Proceedings of ED-MEDIA’95,world conference on educational multimedia and hypermedia (pp. 430–435).Chesapeake, VA, USA: AACE.

Masoodian, M., Apperley, M., & Frederikson, L. (1995). Video support for sharedwork-space interaction: An empirical study. Interacting with Computers, 7(3),237–253.

Matarazzo, G., & Sellen, A. (2000). The value of video in work at a distance: Additionor distraction? Behaviour and Information Technology, 19(5), 339–348.

Millard, N., Coe, T., Gardner, M., Gower, A., Hole, L., & Crowle, S. (2000). The future ofcustomer contact. BT Technology Journal, 18(1), 51–52.

Muller, M.J., Carr, R., Ashworth, C., Diekmann, B., Wharton, C., Eickstaedt, C., &Clonts, J. (1995). Telephone operators as knowledge workers: consultants Whomeet customer needs. In Proceedings of the SIGCHI conference on human factors incomputing system (CHI’95) (pp. 130–137). New York, NY, USA: ACM Press/Addison-Wesley Publishing Co.

Nardi, B., & Whittaker, S. (2001). The place of face-to-face interaction in distributedwork. In P. Hinds & S. Kiesler (Eds.), Distributed work new ways of working acrossdistance using technology. Massachusetts, USA: MIT Press.

Norman, D. (1988). The psychology of everyday things. New York, USA: BasicBooks,Inc.

Ochsman, R. B., & Chapanis, A. (1974). The effect of 10 communication modes on thebehavior of teams during cooperative problems solving. International Journal ofMan-Machine Studies, 6(5), 579–619.

Olsson, A.C. (2007). Understanding and enhancing customer-agent-computerinteraction in customer service settings. Unpublished PhD thesis. Universityof Waikato. Hamilton, New Zealand.

Orvis, K. L., Wisher, R. A., Bonk, C. J., & Olson, T. M. (2002). Communication patternsduring synchronous Web-based military training in problem solving. Computersin Human Behavior, 18(6), 782–795.

Randall, D., & Hughes, J. A. (1995). Sociology CSCW and working with customers. InThomas (Ed.), The social and interactional dimensions of human–computerinterfaces. New York, NY, USA: Cambridge University Press.

Rodden, T., Rogers, Y., Halloran, J., & Taylor, I. (2003). Designing novel interactionalworkspaces to support face to face consultations. In Proceedings of the conferenceon human factors in computing systems (CHI’03) (pp. 57–64). New York, NY, USA:ACM Press.

Rutter, D. R., & Stephenson, G. M. (1977). The role of visual communication insynchronizing conversation. European Journal of Social Psychology, 7(1), 29–37.

Rutter, D. R., Stephenson, G. M., & Dewey, M. E. (1981). Visual communication andthe content and style of conversation. British Journal of Social Psychology, 20(1),41–52.

Scaife, M., Halloran, J., & Rogers, Y. (2002). Let’s work together: Supporting two-party collaborations with new forms of shared interactive representations. InProceedings of COOP’2002, Nice, France, August, (pp. 123–138). Amsterdam,Holland: IOS Press.

Short, J., Williams, E., & Christie, B. (1976). The social psychology oftelecommunications. New York, NY, USA: John Wiley & Sons, Ltd.


Steel, A. (2003). Understanding and enhancing call centre CHHI. In ProceedingsCHI’03 extended abstracts on human factors in computing systems, SESSION:Doctoral consortium submissions (pp. 690–691). New York, NY, USA: ACM Press.

Steel, A., Jones, M., & Apperley, M. (2001). Auditory feedback in call centre CHHI. InProceedings OZCHI 2001: Usability and usefulness for knowledge economies (pp.142–147). Perth, Western Australia, Australia.

Steel, A., Jones, M., & Apperley, M. (2002). The use of auditory feedback in call centrecomputer–human–human interaction. In CHI’02 extended abstracts on humanfactors in computing systems (pp. 68–69). New York, NY, USA: ACM Press.

Suchman, L. A. (1987). Plans and situated actions: The problem of human–machinecommunication. New York, NY, USA: Cambridge University Press.

Tanir, O., & Booth, R. (1999). Call centre simulation in Bell Canada. In Proceedings ofthe 31st conference on winter simulation: simulation – A bridge to the future (pp.1640–1647). New York, NY, USA: ACM Press.

Ten Have, P. (1991). Talk and institution: A reconsideration of the asymmetry ofdoctor-patient interaction. In D. H. Zimmerman & D. Boden (Eds.), Talk andsocial structure: Studies in ethnomethodology and conversation analysis.Cambridge, UK: Polity Press.

Walker, M.A. (1994). Experimentally evaluating communicative strategies: Theeffect of the task. In Proceedings of the twelfth national conference on artificialintelligence, Vol. 1 (pp. 86–93). Menlo Park, CA, USA: American Association forArtificial Intelligence.

Human communication in customer-agent-computer interaction: Face-to-face versus over telephone

Documents

Transcript of Human communication in customer-agent-computer interaction: Face-to-face versus over telephone