INSPIRE is a Web-based system for the support and conduct of negotiation. The primary uses of the system are training and research. Between July 1996 and April 1997, 281 bilateral negotiations were conducted through the system by managers, engineers and students from over 50 countries. INSPIRE has been used at eight universities and training centers. In research it is being used to study cross-cultural differences in decision making and the use of computer support in negotiation. This paper outlines the system, the negotiation methodology embedded in it, and reports the initial results of the experimental study of the impact of culture on Web-based bilateral negotiation.

Keywords: negotiation, international negotiation, cross-cultural study, Internet, World Wide Web, decision support, negotiation support, preferences.

The authors gratefully acknowledge the contribution of each member of the InterNeg team and, in particular, they want to thank John Bowen, Ying-Hueih Chen, Anantha Mohadevan, Rajeev Mohan, Kumudini Ponnudurai, and Ravi Ramsaran for their help and enthusiasm. We also want to thank Dr. Madan Mohan from the Indian Institute of Management Bangalore for his help in organizing and running negotiations. The project has been supported with grants from the Natural Sciences and Engineering Research Council, Social Sciences and Humanities Research Council and the International Institute for Applied Systems Analysis.

Gregory E. Kersten is a Senior Research Scholar with the Decision Analysis and Support Project of IIASA.

Sunil J. Noronha is with the T.J. Watson Research Center, IBM, Yorktown Heights, USA.

1. Introduction

1.1 Negotiation and technology

Computing and communication technologies introduce new and exciting ways of making decisions, engaging in social processes, and also conducting research experiments. These three broad activities are the main focus of the InterNeg project, and this paper which reports the project’s results.

Improved access to data, information, knowledge and expertise allows one to make better informed decisions. Availability of specialized software enables analysis and evaluation of difficult problems to a degree previously only available to experts. Technology allows for media rich communication among individuals and groups across borders, conduct of different transactions, and their engagement in exchanges, including collaboration and negotiation. Finally, technology makes it possible to conduct research experiments across borders, involve subjects from different countries to simultaneously participate in the experiments, and also to employ computation-intensive techniques for the analysis and visualization of data.

The use and application of computer and communication technologies plays two important roles here. One is related to the development of INSPIRE, the first Web-based decision and negotiation support system and also other systems used for negotiation support and evaluation, and for data collection and analysis . The second role pertains to the usage of INSPIRE and its users who come from different cultures. The technology allows the users to employ techniques for the analysis of decision and negotiation and to conduct negotiation over time and space.

The study and experiments are part of the InterNeg project which involves:

In this paper we make an attempt to give light to the differences in axioms, goals and objectives leading to different processes and outcomes. There are several key differences between this study and other studies

• the use of computer and communication technologies to observe the process of negotiation in a controlled setting,
• negotiations can be conducted anonymously thus the cultural bias may be reduced,
• negotiators have access to decision and negotiation support tools,
• the negotiation case allows for specification of subjective preferences among issues and options, and
• the negotiations may be conducted over several weeks with or without imposed deadlines.

Two systems are currently available in InterNeg: INSPIRE and INSS. In this paper we concentrate on the use of the INSPIRE system which has been specifically developed to study negotiation processes and negotiators ’ behavior. This system allows for a large scale systematic study of cultural differences in negotiation, which were previously nonexistent.

INSPIRE experiments are still being conducted and we expect that the number of participants will significantly increase in the future. For now, although people from many countries used the system, most countries are represented by only few negotiators. We are also continuing work on the software that processes negotiation records which are automatically generated during negotiations. All these reasons contribute to our reluctance to present results in the typical form of hypothesis and proves. Instead, we present here summarized data and point out what we think are interesting and worth further studying phenomena. Although the paper reports on an ongoing research program we hope that the uniqueness of the study and the results already collected justifies it.

The paper has 5 more sections. In the rest of this section an overview of earlier studies and their results are given. The methodology underlying negotiations via INSPIRE is presented in Section 2 and the system itself is outlined in Section 3. The type of data generated by the INSPIRE negotiations and programs that are used to process it are discussed in Section 4. Section 5 presents results of 281 bilateral negotiations and focuses on the characteristics of negotiators from five countries. Discussion of the results and future work conclude the paper.

There are four main types of studies of negotiation and culture:

1. questionnaires,
2. experiments,
3. case studies, and
4. experts’ opinions.

The first type of study involves the analysis of usually a large number of questionnaires asking people about their perceptions, reactions to simple situations, values, and opinions. The most well known study is the Hofstede-IBM with 116,000 questionnaires containing the values of employees of IBM in 72 countries . In his analysis, Hofstede found four culture-defining main dimensions, namely, power distance, individualism, masculinity, and uncertainty avoidance.

The second type involves the conduction of the same experiments in several countries, typically among university students and participants of executive courses. Questionnaires are also used in this type of study, but they pertain to the subjects perception of the game and the outcomes. Several well known experiments of this kind were conducted and reported (missing one more and important). The experiments are short and take an hour or two. Often Kelley’s negotiation simulation is used . The simulation involves bargaining for prices of three commodities and lasts no more than one hour; thus it is suitable for a classroom environment. It has often been used in the analysis of bargaining and its outcomes .

Case studies involve the observation and analysis of real-life cases of international negotiations that take weeks or months to complete. Examples here include the now classical discussion of cross-cultural negotiations by Gulliver , several works about the Law of the See negotiation (see, for example, ), the Cuban missile crises, Camp David negotiations (Raiffa 1982), and so on. The case method is advocated by the Project on International Negotiations (PIN) whose members collected and published many cases (see for example, .

The forth type involves negotiators’ retrospection, information and experiences contained in their own writings (Fisher 1980; Cohen 1991). Definitely a lot can be learned from the analysis of experts’ opinions and experiences. However, typically those negotiations are atypical and involve very high stakes, are important to countries or powerful organizations. Moreover, the opinions are provided by highly trained professional negotiators and diplomats who have their own trans-national and distinct culture .

Until recently it is only the second and third type of study that allow for the analysis and assessment of negotiations typical for real-life negotiations. For practical reasons, the experiments did not involve negotiations they extend beyond one or two hours. Another characteristics of the experiments is the subjects’ knowledge of each other and their close proximity. Further, the subjects typically were from the same culture and negotiation did not involve any cross-cultural communication.

Cultural studies were done on the basis of negotiations conducted in culture X with those conducted in culture Y . Thus, very little can be said about international and cross-cultural negotiations. An exception is the study in which the intra-cultural negotiation are compared with cross-cultural. This study involved, in the cross-cultural experiment, 30 negotiations between Americans and Japanese and 26 between Anglophone and Francophone Canadians .

Experimental studies make it possible to analyze, assess and possibly measure specific mechanisms and methods used in the process, and attitudes and perceptions of the subjects. This is achieved at a cost of a highly stylized and unrealistic negotiations process and its setting. Another difference between the traditional experiments and other studies is in the importance of negotiations. Experiments typically deal with fairly common negotiations, not critical to a country, organization or an individual. Other studies often focus on negotiations involving politicians and experts.

The changes that societies, groups and individuals face alike involve the relative shrinking of the world, globalization of the economies, internationalization of medium and even small companies, and -- most importantly -- quickly increasing and rich communication among people from every corner of the world. Some corner stores may now be thousands of miles away from the buyer and yet retain the "corner store feeling" though now this store becomes virtual. This has an obvious impact on the conduct of bargaining negotiation. Deals will be made by people who do not know and see each other.

There has been many studies of international, cross- and intra-cultural negotiations. We mention here only a few which are either of the same type as this study, or which are of direct significance to the results presented here or which are being currently analyzed.

An interesting study involved an examination of the bargaining behavior of children from India, Argentina and the US which found -- among others -- that Indian bargainers were more competitive than Americans and Argentineans . This observation, though with respect to professionals and more complex negotiation, is confirmed in the present study.

Graham observed that negotiators change their behavior significantly depending if they are engaged in cross-cultural negotiations as opposed to intra-cultural . Another study reports that Americans were more satisfied, Japanese achieved lower profits and higher interpersonal attraction, French Canadians were more cooperative, and English Canadians achieved lower profit and spent more time negotiating in their cross-cultural than intra-cultural negotiations .

An intra-cultural study involved a series of experiments with students from four countries, Israel, Japan, former Yugoslavia, and the US . Negotiations were of the pure bargaining type with one issue being the price. The results suggest that there are statistically significant cultural differences in the size of offers (low-high), percentage of rejected offers and thus in Pareto-inefficient outcomes. Further, they found that the between-country differences in the outcomes became smaller as the bargainers gained experience.

Negotiations involve communication. Yet most studies concentrate on reported perceptions of negotiation processes and outcomes in questionnaires and worksheets and ignore the vital role of communication (e.g., ; Ting-Toomey et al., 1991). While in this paper we also mostly report on negotiators perceptions, INSPIRE negotiations allow for free text communication and we have collected and will analyze all the messages. The results presented in this paper confirm that "Culture influences negotiation through its effects on communication" , but also suggest a broader, than communication, scope of these influences.

Experiments with Taiwanese and American subjects, in which the communication in bargaining situations was analyzed, showed a significant difference between negotiators’ perceptions and actual interaction patterns . Drake observes that despite cultural differences, the same cultural differences may not emerge in face-to-face interactions with negotiators from different cultures. If this is the case then the INSPIRE negotiation may allow one to observe what cultural differences, and in which circumstances, emerge in anonymous negotiation. It may also provide information whether negotiators significantly change their behavior when moving from intra- to cross-cultural negotiations .

The negotiation problem involves two companies: Itex Manufacturing, a producer of bicycle parts and Cypress Cycles which builds bicycles. In writing the case an effort has been made to make it as much as possible ‘culture neutral’ . Furthermore, the case was to describe a negotiation situation with which users from almost any country are familiar without any additional explanation. As the users’ language proficiency might be low the case is fairly simple and well structured. It case description fits one and a half pages.

Cypress Cycles, an established manufacturer of high quality mountain bikes, is launching a new line of bikes and requires a type of component that its current suppliers cannot provide. Their first serious discussions for the supply of these components are being held with Itex Manufacturing. Itex is seeking to increase its share of the component market and would like to have the prestige that would come with supplying Cypress, should a profitable contract be concluded. There are four issues that both sides have to discuss and they are the price of the components, delivery times, payment arrangements and terms for the return of defective parts. The negotiators were not given the issue priorities, thus they had to decide if, say, price was more important than delivery time and the specific trade-off values between issues.

For each issue there is a set of options, i.e., issue values. Altogether, there are 180 complete and different potential offers (alternatives) that contain all four issues.

Both parties are presented with their side of the case, told that they are to represent Itex and Cypress respectively, and that their companies are interested in achieving a compromise. However, they are also informed that there are other suppliers and buyers so that a breakdown in negotiations is possible if they cannot reach a good deal. There is no further specification as to what indicates a good deal.

Each side is given a clear indication as to the desirability of the options (issue values) but only in a sense of the direction and not specific trade-off values. That is, it stated in the Itex’s case description that a higher price is better for Itex, the seller. Similar indications are given with respect to other issues. An example of the wording for the issue Returns, which describes the return policy of defective parts, is given below.

1. Full price on all returned parts. Parts returned at Cypress' option. Itex pays shipping on all returned goods 75% refund.
2. Five percent spoilage allowed. If more than five percent of a shipment is unacceptable the whole shipment is returned for a 75% refund.
3. Ten percent spoilage allowed. If more than 10% of a shipment is unacceptable the whole shipment is returned for a 75% refund.

Specification of the preference direction and not value allows negotiators to establish their own priorities within each issue. Furthermore, the negotiators partial utilities (part-worths) may be linear as well as non-linear. Because there is no mechanism enforcing the preference direction, some negotiators did not follow the preference direction literally. Rather, they assigned the maximum partial utility to one of the intermediate and not extreme options. For example, they assigned a higher reference to the second option of the Return issue and not to the first option.

The literature suggests three sequential phases in studying business negotiations: an antecedent phase, a concurrent phase, and a consequent phase (Graham, Mintu et al. 1994). These three phases roughly correspond to three phases of the negotiation: pre-negotiation analysis, conduct of negotiation, and post-settlement analysis.

The pre-negotiation phase involves analysis of the situation, problem and opponent, formulation of preferences, reservation levels, BATNA, and strategy. Data on the negotiation problem, negotiators’ characteristics, including their preferences, and situational constraints are considered within the antecedent phase . In the INSPIRE negotiation two main instruments are used to collect the data:

1. forms used to elicit preferences and construct negotiator’s utility function, and
2. a pre-negotiation questionnaire which every negotiator has to fill in after her/his utility function has been constructed and before negotiation can begin.

The conduct of negotiation phase involves exchanges of messages and offers. Offers comprise the negotiated issues and their values, e.g., one of the three value of the returns policy given above. The negotiation is not sequential but parallel; an offer comprises all the negotiated issues. Participants may submit the same offer many times, or keep the option of an issue unchanged, but each submitted offer contains an option for each issue.

The negotiation ends when a compromise has been achieved or one of the parties terminates the process and informs their opponent. The concurrent research phase corresponds to the conduct of the negotiation and it comprises process-related variables, such as the strategies and behaviors used by negotiators , changes in the negotiation problem and negotiators’ perceptions, and the dynamics of negotiations .

The post-settlement analysis phase may be static and involve only the evaluation of the negotiation outcomes generated by, and after, the negotiation activity . These outcomes include the information about the compromise and the negotiators’ satisfaction. The analysis thus focuses on the evaluation of variables describing outcomes. However, if the negotiation allows for the analysis and improvement of the compromise efficiency then the post-settlement phase may also involve continuation of negotiation in order to improve the compromise. The INSPIRE users have the possibility to improve inefficient compromises.

The post-settlement phase ends with filling in the post-negotiation questionnaire which, however, is not mandatory. A user may log out from the system or -- upon filling in the questionnaire -- is directed to multiple negotiation resources, handouts, systems which are available on the InterNeg site, but about which users are not informed during negotiation.

We assume that users posses a basic knowledge of English and are able to use Netscape. Although the system has been used, among others, by three groups of English as a Second Language (ESL) students, users did not experience difficulties understanding the case. ESL students at Carleton University, some with no computer exposure, were given one hour hands-on tutorial and were guided during their first session with INSPIRE. They had no difficulties in using the system in the subsequent sessions.

The participants do not get any financial reward. Many of the participants used INSPIRE as part of their course, however, none of those considered here were evaluated (marked) on the basis of their performance.

INSPIRE is a support system based on analytical models rooted in decision and negotiation analysis . The system and its architecture have been described in detail elsewhere . A short outline follows.

INSPIRE represents the value of negotiation-related constructs---issues, options, and offers (packages)---to each negotiator by means of utility functions. This representation forms the basis of a scoring scheme that enables negotiators to make easy comparisons between offers and counteroffers and judge the significance of a concession.

The technique currently implemented for the construction of utility functions is based on conjoint analysis, in which the utility of a given package is determined from the user's preference orderings over a set of factorially designed packages, . A hybrid (compositional as well as decompositional) approach is used and it comprises three steps:

1. The user evaluates the relative importance of the issues to be negotiated. The rating assigned to each issue is viewed as a component of the total utility of a package. The utility component of each issue is assumed to be independent of the other issues, i.e., any possible interactions are assumed to be insignificant. Therefore the utility components are simply added together to form the total utility function and this is called composition.
2. The user evaluates the relative importance of each issue's options. The rating of each option constitutes the utility component of an issue when that particular option is the one that's present in a package.
3. The user makes a comparative evaluation of several complete packages selected by INSPIRE, viewing each package as a whole. This is the decompositional step. The total utility of a package is decomposed into constituent option utilities using an additive model:

There is a large number of packages that could be presented, and we need some way of selectively presenting just a few packages for the user to rate, yet obtain reliable utility values. This is a problem in the design of fractional factorial experiments. One of the most compact and effect designs is the orthogonal design, in which the packages are chosen such that the X matrix is orthogonal. INSPIRE uses the information obtained in the issue and option ratings steps to select the set of orthogonal packages presented to the user for the package rating step. Given the ratings for these packages, the weights u_ij are computed so as to minimize the error terms using linear regression.

Since the utility of every possible option is considered explicitly, the utility function for a given issue can be nonlinear. This is an advantage since people usually do not have linear utilities as they traverse a given range of values. By default, issues are assumed to have "discrete" options, i.e., only a small number of explicitly listed options are considered to be meaningful as outcomes of the issue. These are also called salient options. However, some issues can also be "continuous" in the sense that any intermediate value can be meaningful. In such cases, the utility function within an issue is assumed to be piece-wise linear, i.e., linear interpolation (or extrapolation) is used to compute the utility of intermediate points between salient options.

Evaluating the utility function with respect to the combination of options that comprise an offer provides a numerical estimate of the goodness of the offer. INSPIRE uses this in several ways: it generates graphics plotting the score versus time, thus enabling the negotiators to understand at a glance the history of concessions that have occurred; and it uses the scores to suggest possible improvements on any compromise that is reached. In the interest of the study INSPIRE currently refrains from making full use of the support possibilities suggested by utility functions, e.g., recommending or shortlisting good counteroffers during the conduct phase, and preventing the user from composing non-optimal offers during the post-settlement phase, since strong guidance from the system could mute the differences in natural behavior.

A negotiator’s preference information is never revealed to his or her counterpart, or anyone else, as in real negotiations. This implies that evaluation of a compromise that has been accepted by both parties must be done independently for each side, using the corresponding utility function; interpersonal comparisons of utility are never performed. A compromise is considered Pareto-optimal or efficient if it cannot be "improved," i.e., if there does not exist any other package with a higher score for one party and an equal or higher score for the other party - measured with their respective utility functions.

Whenever a compromise is reached, INSPIRE determines whether it is Pareto-optimal. If it is, the negotiation ends; else, INSPIRE computes the set of all packages that dominate the compromise, and presents five that span the spectrum of scores they represent. The user is given the choice to select one of them or construct a new offer altogether, or terminate the negotiation and stay with the inefficient compromise that has been achieved. Note that this computation requires INSPIRE to make simultaneous use of both utility functions, without actually revealing them to the negotiators. In other words, the character of the post-settlement phase is subtly different with respect to all other activities performed by INSPIRE: it requires a controlled and limited sharing of the two negotiators’ preference information in the form of the results reported by a trusted third party (INSPIRE), whereas all other processing requires access only to one’s own preference information. While this has not been of any significance in the academic contexts of the study, it may be an issue in sensitive real-world negotiations where third-party trustworthiness is an issue; the likely result will be reduced usage of such support features.

The traditional view of a negotiation (or group) support system is that of a desktop application: each user has one copy of the software on their personal computer, which communicates with the other users' copies over a network (typically a LAN), usually in synchronous mode (i.e., with both parties simultaneously logged on). Figure 1 depicts how INSPIRE’s process model, conceptualized as a negotiation support system, has been translated into its implementation structure as a Web application. The system uses the client/server model of distributed systems to partition the main components.

Figure 1. Negotiation support and its implementation in INSPIRE

Three major factors affect the system’s design:

1. We wish to enable users with nothing more than a Web browser and an Internet connection to avail of INSPIRE’s services. This implies a tremendous degree of portability and gives the researchers access to users in remote countries with minimal computing resources.
2. Current trends in net-centric computing are towards pay-per-use software: programs reside at their developers' home sites and are automatically downloaded and executed whenever the user needs a particular piece of functionality. This induces a tendency towards an architecture in which the server (INSPIRE’s home site) plays a central role, regardless of the structure of communication needs.
3. Since one of the INSPIRE’s primary goals is to observe and log user activities as completely as possible for the cross-cultural study, and since it is difficult to monitor actions on the user's host machine, it is desirable that all nontrivial activities be conducted through the INSPIRE site.

One other factor that significantly influenced the design is the fact that the negotiations supported are asynchronous: since the two parties negotiating with each other typically reside in far away countries with different time zones, it is rare for both sides to be simultaneously logged on. Therefore, INSPIRE is designed to interact independently with each user, saving the state resulting from each user's actions in a form that can be retrieved when the counterpart logs on some time later.

INSPIRE presents itself to the user via Web pages containing, apart from paragraphs of hypertext providing context-specific guidance, forms for user input and graphics for visualization. Each page is in essence a snapshot of the situation or context at a given moment in the negotiation process; the contents of the page are of course dynamically generated since they differ for each negotiation. Each page provides the user with many prompts (hyperlinks or buttons) for the various things he or she can possibly do at that point. Since some of the page structure can be prespecified, INSPIRE’s architecture has been built around two major components: the frontend, comprising skeletal HTML/JavaScript pages implementing contextual structure, prespecified instructions and choices, and the engine, a set of large C++ programs that implement the negotiation methodology and provide functionality such as messaging, user authentication, session management, etc.

All of INSPIRE’s implementation is object-oriented, and each piece of functionality is provided by a group of object classes that is loosely coupled with the rest of the system. Therefore each of them has been implemented fairly independently of the others. An effort was made to construct software objects that can be used and reused in different conditions and in different configurations. Moreover, we aimed at a system which in future can be expanded and use new and additional components. This has led to a more powerful cousin of INSPIRE called INSS, supporting sequential negotiation and continuous issues; however, the latter has not been used within our cross-cultural study and will not be discussed further.

The INSPIRE system provides two sources of data, which together describe the entire negotiation as well as the negotiators themselves:

1. two questionnaires that are filled on-line by each negotiator, and
2. the complete computer records of the negotiation.

The first questionnaire is filled out during the early preparation phase of negotiation (the ‘pre-negotiation questionnaire’), and the other after the negotiation is terminated (‘post-negotiation questionnaire’). In brief, these questionnaires try to capture background information about the negotiators that would otherwise be unavailable, given that they are typically unknown people coming in over the Internet. They also directly elicit perceptions and judgmental information about each other and the negotiation environment.

The pre-negotiation questionnaire contains 16 questions abut the user’s background, including age, countries of birth and residence, self-evaluation of negotiation experience, level of Internet use, prior knowledge (if any) about the counterpart, and expectations about the nature of the forthcoming negotiation and the compromise that will be reached. The user is typically required to specify the latter judgments on a scale of five levels whose extremes are labeled, e.g., from ‘Very friendly’ to ‘Very hostile.’ Two additional questions are about the difficulty of INSPIRE’s preference elicitation mechanisms which are later used to provide users with the subjective values of alternatives. The system does not allow users to proceed to the second phase of negotiation (exchange of offers) until they fill in this questionnaire.

The post-negotiation questionnaire contains 40 questions, including a few open-ended requests for comments about the system’s features and potential. The questions are about the system and its features (17 questions), the agreement reached (2), the process (4), the negotiator and the opponent (17). The questions about negotiator and the opponent are grouped together, because typically the negotiator does not know her/his opponent. Thus, almost any question about the opponent in fact describes the negotiator’s perception and not reality. Examples of such questions are: whether the negotiator found the opponent informative, persuasive, honest, etc. In Adler’s terms, these questions measure the opponent’s attractiveness’ and the negotiators’ ‘problem solving approach’ (Adler, 1989).

INSPIRE’s history recording mechanism logs each negotiator’s activities in detail and provides the complete computer records of the negotiation. This includes information about the negotiator’s use of messaging, visualization, and analytical tools. The negotiation database contains all the objects created and exchanged during the negotiation, including the offers and messages composed, along with time-stamps. In essence, these logs provide two qualitatively different types of data: direct inputs from the negotiators (e.g., the transcripts of their online conversation, their value system as represented by their preference ratings, and the composition of their offer packages), and observed data (e.g., the timing and frequencies of messages and offers, the sequence of scores on the offers and compromises, and the gains achieved during the post-settlement phase).

While it is possible, and fascinating, to study the transcripts of each negotiation individually and analyze the strategies used, for the purposes of our present study the raw data described above has to be transformed into statistics across the entire negotiation database. Further, even within a given negotiation, there are many derived measures of interest, e.g., the mean time between counteroffers, the score on the expected compromise (obtained by applying the negotiator’s utility function to the expected compromise package elicited in the pre-negotiation questionnaire), the difference between the expected and achieved compromise scores, the concession pattern, and comparisons against the corresponding measures for the counterpart.

Much of this computation has been automated through another INSPIRE component called INtoSPSS; the name is an artifact of its original purpose, to massage INSPIRE data into a form suitable for input to the SPSS statistical analysis package. This module shares the bulk of its code with the rest of the INSPIRE engine---and therefore has access to the complete data---but differs in implementation in one significant way: it has been designed to run off-line, not on the Web. Further, its data structures have been designed to enable easy insertion and coding of new derived measures; typically within 3 to 8 lines of C++ each. In its default mode, the program currently presents all the data in a standard tabular form that can be fed directly into SPSS. Other modes are currently available to provide an output format suitable for S-Plus, and to print out the unstructured answers to open-ended questions from the post-questionnaire. The latter is particularly useful for research queries such as "summarize all user comments about system features that were not found useful."

Deciding which derived measures to compute and analyze is not an easy problem; there are too many possible measures, and all of them suggest extremely fascinating research hypotheses (see the following section on structuring the data). Moreover, the researchers’ interests lean towards the discovery of new, interesting and significant patterns in negotiation behavior and tool usage, rather than confirmation of preconceived cultural hypotheses. In other words, the need is for data exploration tools supporting visualization, direct manipulation and data mining. This need has been addressed to a limited extent through programs that we have developed in S-Plus and SPSS, and we are currently exploring other tools such as rough sets. (Pawlak, 1991).

Around 70 measures from each negotiation are currently being analyzed via the tools just described. Each of these variables (factors) constitutes a column in the INtoSPSS output table. Each negotiation contributes two rows to the table, one for each negotiator. In general, each of the 70 variables can be cross-correlated and regressed with one or more of the others in order to examine the implied hypothesis that the given group of variables influence each other. In order to enhance clarity of understanding, it helps to structure these variables a priori using intuitive judgments of causal relationships. Therefore we categorize the measures broadly into independent variables and dependent variables, such that we are primarily concerned with the effect of the former on the latter. Note that these are strictly judgments and often the direction of causality is far from obvious. Moreover, many variables are intermediate in the sense that they are causally influenced by some independent variables, and in turn influence the value of some dependent variables. For example, the number of offers made by a negotiator is influenced by the bargaining norms in her/his culture, and influences the likelihood of reaching agreement with a counterpart from another culture. The following classification therefore only helps identify and focus on the core variables, namely those which most clearly belong to the two extreme categories.

These are factors that are largely controlled by the researchers and are not unaffected by most other variables in the study. They fall into three sub-categories: subject, task, and system variables.

Subject variables are those which describe the negotiators and some of the most important variables for our study, namely cultural markers, fall into this category. Country of birth and country of residence are the two cultural markers most heavily studied; for a discussion see Section 5.1. These are often confounded by other key subject variables such as age, gender, negotiation experience, and language difficulty.

Task variables relate to the problem undertaken by the subjects; these center around the negotiation case. For example, cultural cues in the case description, the domain of the case (strategic, marketing, etc.), the complexity of the case, and deadlines can potentially have significant effects on the observed results of negotiation. For this paper, task variables are not particularly important because they have all been kept constant for the entire study to date. However, varying factors such as team size and multilateral negotiation cases hold the promise of very interesting future studies.

System variables represent the presence or absence of specific support features in INSPIRE. For example, forcing users to go through the initial preference ranking steps can make them more aware of their priorities and alternatives and influence their subsequent negotiation behavior, than if the preference elicitation step were skipped. The scores provided on each package by INSPIRE are another system attribute that can have significant impact on behavior and outcome since it reduces attention with respect to individual issues. The display of various kinds of negotiation visualization aids (e.g., negotiation dance graphs), decision analytic tools (e.g., utility construction methods), the ability to compose offers via structured menus instead of free text, the ability to transfer complex objects during communication (e.g., documents in support of one’s argument), and general ease of use factors of system variables that are controlled in INSPIRE.

Dependent variables of interest in this study are any direct measures of the ultimate effectiveness of decision making, negotiation style, etc. They can be classified into three categories: measures of the goodness of the negotiation’s outcome, the negotiation process, and the system’s effectiveness.

Some variables representing good decision outcomes are objective and easy to measure, e.g., whether an agreement was reached at all, the level of achievement on each issue in the final compromise, and whether the compromise was Pareto-optimal. Others are subjective, but can nevertheless be elicited successfully, e.g., the goodness of the final compromise as scored by each negotiator’s own value system (utility function). Still others are quite hard to elicit, e.g., the quality of the relationship (including goodwill) established at the end of the negotiation, which has long-term implications, and feelings of satisfaction with the outcome. Variables in the last category are usually measured through the post-negotiation questionnaire.

Similarly, variables measuring the goodness of the decision process that has been completed are sometimes objective, e.g., process efficiency as measured by the time to completion, the number and size of messages and offers; whether post-settlement negotiation was conducted to improve the compromise, etc. Subjective process variables include posterior judgments of satisfaction with the process (which is important because high satisfaction implies that the cognitive complexity of the system was at a tolerable level), recognition of cultural differences, feelings of being better prepared, etc.

Intermediate variables are a complex set of measures that link both the dependent and independent variables; achieving some sort of causal structure on this set would amount to creating a very useful negotiation factor model.

One class of intermediate variables is the initial ‘psychological baggage’ or ‘habitual domains’ that the negotiators bring to the table; this is dependent on the cultural background of the negotiators and impacts the process and outcome of negotiations. For example, in cultures where competitive bargaining is the norm, there may be greater expectation of hostility than in cultures where relationship-building is the basis for negotiation. Initial expectations about the final agreement, and the value system of the negotiators (as measured by the preference rankings) on various criteria are other examples in this category.

A second major class of intermediate variables relates to the behavior of the negotiators during the negotiation. For example people from high-context cultures may spend a greater amount of effort ‘creating context’ during the initial phase of the negotiation -- they may give up anonymity more quickly by introducing themselves, they may have a lower proportion of offers without attached messages, and they might be more verbose in an attempt to overcome the barriers of online communication. Linguistic patterns, message and concession patterns, timing and choice of offers (strategies), surprises created or encountered and disregard for certain system features (such as Pareto-optimal analysis) are other examples in this category.

A third major class involves the perceptions of the negotiators after the negotiation. Perceived level of control over the negotiation, and judgments about and empathy achieved with the counterpart are examples in this category. While these variables themselves are measures of overall performance, dependent on many prior factors, they in turn influence the ultimate process and outcome measures; e.g., a negotiator who felt in control may perceive herself as having done well, even though more objective measures of outcome might prove otherwise.

There are many other types of potentially confounding intermediate variables which we have briefly considered but disregarded for the purposes of this study, e.g., organizational factors and variation in decision making styles.

Between September 1996 and May 1, 1997, 281 bilateral negotiations have been initiated with INSPIRE. 86 of them are considered unusable for analysis mainly because users, after the initial request, did not initiate the negotiation. The remaining 195 negotiations are considered for further analysis.

There are two main types of INSPIRE users:

1. participants of university courses and seminars, and
2. Web surfers.

It is mainly the Web surfers who have dropped out and their negotiations were unusable.

The literature on previous studies indicates that the culture of participants is assumed to be of the country in which the experiments are conducted. Countries with high immigrant populations like Canada and US and the internationalization of the university education, make such an assumption doubtful. Therefore, we have asked participants for information about their country of birth, residence, countries visited for a period of more than two months, and the first language learned. In this paper it is assumed that the national culture is defined by the following four-step procedure:

1. the country of residence and language if they coincide, otherwise by,
2. the country of birth and language if they coincide, otherwise by,
3. the country of the first language learned if not English, Portuguese, Spanish, otherwise by the country of residence, and finally by
4. the country of birth.

The proposed procedure for defining culture is not without its drawbacks. It allows to identify people who are in a country of ethnically different culture as, for example a Pole in Canada or an Indian in the US It will, however, fail to distinguish an Irishman who studies in Canada (if he/she states that Canada is her/his country of residence). Furthermore, it cannot distinguish between people who have emigrated to a country half a century ago from those who are there for several months. Nevertheless, it is an improvement over the classification based solely on the country of residence with a large population of foreigners (e.g., foreign students) or immigrants. An indication of the cultural diversity and the role of migration is that if we consider participants’ country of birth INSPIRE was used by people from 55 countries, but if the country of residence is considered there are only 22 different countries.

This migration as well as the participation of Web surfers contributes to the fact that most countries are represented by only a few people in the data. In this paper, we take a "20 participants" cut-off point, that is countries with at least 20 users are analyzed. Using this criterion and the criterion of culture defined on the residence, birth and language we have five countries: Canada, China, Finland, India and US Together, they comprise 58% (227) of the total number of negotiators.

Table 1.

Implications of two definitions of culture

The implications of the two definitions of culture are shown in Table 1. The change in the definition leads to a significant change in the five country negotiators’ population. This is because many foreign students who study in Canada and the US are taken into account when culture is defined by the country of residence. The change in definition also changes the populations within countries. This is primarily indicative to Canada and China, as many Chinese students from the Canadian universities used INSPIRE. We want to stress here that we make no attempt to provide an ‘objective’ definition of culture nor try to partake in the discussion about the cultural adaptation and change of immigrants and foreign students.

The data about the users of the INSPIRE system is collected in the pre-negotiation questionnaire. The vast majority of the overall population are university students in Canada (University of New Brunswick, Carleton University, and McMaster University), China (Hong Kong Baptist University and Carleton University), and Finland (Abo Academy and Helsinki School of Economics). In India, the negotiations are carried out by managers and engineers during their several-weeks long executive courses at the Indian Academy of Management in Bangalore; the US negotiations were carried out by researchers from several institutions and a group of graduate students from MIT.

The mean and standard deviation of the variables describing negotiators’ is given in Table 2. As one may expect, younger, on average, negotiators consider themselves less experienced than the older ones. However, the correlation is weak; the 2-tailed Pearson correlation between age and negotiation experience is .329 at the 0.01 level.

Negotiators’ access and use of the Web is high, except for Indians who at present experience difficulties with access and speed but who unanimously expect a significant increase in access to the Web. Interestingly enough, while Canadians and Chinese access to the Web is very similar at present, their expectations are quite different as only 38% of the former versus 80 of the latter expect increased access.

For the majority of participants the INSPIRE system is the first DSS/NSS they have used. The exception are Finns who are students from MIS/DSS courses and thus were exposed to other types of computer-based decision support.

The negotiation participants were asked whether they know their opponent or the country she/he is from. Although some of them replied positively, in most cases they mistakenly assumed that the opponent is from the same course and/or from North America. This is confirmed by users’ statements about their opponents and their country them made after the completion of negotiations (see Section 5.6).

The total number of bilateral negotiations is 195 with 390 negotiators. To indicate the buyer’s (Cypress) and the seller’s (Itex) culture, in Table 3, we present the number of negotiators representing each company. It follows from the data that distribution of negotiators into buyers and sellers is not equal and, for example, many more Finns were sellers than buyers and the reverse is true for Indians and Chinese. We plan to equalize this in the future negotiations.

Table 2.

Negotiators