Training in negotiation using artificial intelligence-based simulation software

Dr. Francisco Ortega, Project Engineering Area, University of Oviedo, Spain
Dr. Joaquín Ordieres, Project Engineering Area, University of Rioja, Spain
Mr. Javier de Cos, Project Engineering Area, University of Oviedo, Spain
Dr.Valeriano Álvarez, Project Engineering Area, University of Oviedo, Spain

Introduction

It is clear that project management is a specific field that requires specific formation. The advances in communications and computing make possible the use of new technologies in every field. Formation in project management is not an exception.

Simulation games are a new way of training, based in the modeling of a process (in this case the project management or the negotiation process) and its use through a computerized interface. With this approach it is possible to create virtual scenarios different for every user, to introduce variations randomly and to act accordingly to user behavior.

Negotiation is one of the aspects of management where simulation can provide an important training tool. The project engineering area from University of Oviedo, in collaboration with University of La Rioja has developed a tool to train project managers in the field of negotiation in an autonomous but interactive way. The tool, called ESNEF, presents several cases to the user and it acts following some criteria derived of a fuzzy logic network. Offer is not 100 percent good or bad as there is a continuous output function of multiple aspects. So it is necessary to implement a non-conventional logic system more complex than Boole logic guided by absolute values. Fuzzy logic has been selected as the best fit technique to combine complex reasoning and playability.

Examples are based in expropriation problems. A project requires the acquisition of a property with the intervention of the government. This case was selected as it is well known in civil engineering and present several different roles for the user as there are three parts involved: proprietary, public administration, and beneficiary.

System has multimedia capabilities and it is prepared for distance learning, as it will be accessed via web.

This is a didactic tool, so user will be guided but also informed of the steps of his opponent. Game is interactive, final success will depend of the way he will use the available information: Playing user learns to play.

Simulation Games in Project Management

Simulation games have extended from other fields, mostly economics, to related disciplines as project management. From 1990 some tools for training in general project management have been developed, although they are not extended yet. PROSIGA (Cano 1999), PROSIM from Italy, SESAM from University of Stuttgart, or SPS from Arizona State University are perhaps the most extended.

The simplest form of negotiation is presented in the Prisoner’s Dilemma (Poundstome 1993). Although it is more a problem of group dynamic than a negotiation, it is a first approach of interaction between two parts with different objectives. It is easy to find online versions of that problem (for example in http://netrunners.mur.csu.edu.au/%7Eosprey/prisoner.html), generally developed with the tit for tat strategy (one player imitates the other).

More advanced tools where designed by the Austrian IIASA institute. INSPIRE is a web-based system (http://iiasa.ac.at/Research/DAS/inspire) for bilateral negotiations. The system has been used at eight universities in executive training, graduate and undergraduate courses. It has evolved to INSS. In contrast to INSPIRE, INSS allows for adding additional techniques and support tools, e.g., for preference elicitation and representation, construction and update of utility functions, determination of efficient solutions, contextual analysis of strategies and tactics, and qualitative determination of utilities or other parties. Several such tools will be built and deployed in 1998. There is not information about advances from that point.

Finally there are other tools designed for mediation and conflict solution in negotiation, some of them very interesting. Smartsettle aims to accelerate the negotiation process and put decision-makers in control of a process that finds the best possible solutions. It allows parties to participate in the most appropriate combinations of face-to-face meetings, conference calls, and online exchanges in order to quickly find a fair and efficient resolution. One Accord, an interactive computer-assisted negotiation support based in the ICANS patented algorithms, Art of Negotiating®, Negotiator Pro, or WinWin are other tools, in those cases without web interface.

Exhibit 1. Examples of Membership Function

Examples of Membership Function

Techniques in Negotiation Simulation

Most simulation games use linear programming as motor of the models. In other cases binary trees or a discrete set of empiric rules are the base of the system. Those technologies present several limitations. For example the use of predefined rules implies two problems:

• Scenarios are repeated, even if there is a big database of rules.

• User learns the behavior of the system, so results are improved with successive playing, but that improvement is not due to real learning of negotiation techniques but to adaptation to system behavior.

Mathematical formulation based in specifically developed algorithms presents similar problems. If formulas are not specially complicated, user is capable to learn the behavior of the computer in such a way as a sensibility analysis. Although the system could work perfectly for different cases every time or with different users (for example in real conflict mediation as in SmartSettle), it is not easy to be applied to a teaching process where the access to the same problem must be repeated consecutively by the same user, but circumstances must be different in order to identify the evolution of the user as he applies more theoretical knowledge to his decisions.

The introduction of heuristic algorithms, as variations of the tit for tat algorithm, could be used in the case of negotiation, but it produces a very soft system, with inconsistent response and far from the reality. Other AI algorithms could be also applied, as the MINIMAX graphs, but decision trees present very limited performance due to the limited number of rules and the difficulty to build an adequate inference motor. Also “frames” from Minsky and specially Scripts, are interesting approaches, but too simple. It is necessary to consider that the problem will be generally non-linear and solutions must be capable to deal with complex environments.

Fuzzy Logic

In 1965, Zadeh published the first paper on a novel way of characterizing non-probabilistic uncertainties, which he called fuzzy sets (Zadeh 1965). Fuzzy sets are a powerful tool to represent the approximation intrinsic in perception and reasoning either in human beings and animals. Their applications, which are multidisciplinary in nature, includes automatic control, consumer electronics, signal processing, time-series prediction, information retrieval, database management, computer vision, data classification, decision-making, and so on.

Fuzzy logic imitates human reasoning as it considers crisp boundaries instead of binary logic; that is the decision whether an element belongs or not to a set is gradual and not binary. For example, a classical set A can be expressed as

img

where there is a clear, unambiguous boundary point 6 such that if x is greater than this number, then x belongs to the set A, otherwise x not belong to this set. In fuzzy logic the transition from “belonging to a set” to “not belonging to a set” is gradual, and this smooth transition is characterized by membership functions that give fuzzy sets flexibility in modeling commonly used linguistic expressions, such as “the water is hot” or “the temperature is high.”

If X is a collection of objects denoted generically by x, then a fuzzy set A in X is defined as a set of ordered pairs:

img

where μΑ(x) is the membership function of x in A, and it maps each element of X to a continuous membership grade between 0 and 1. When membership value is close to the value 1, it means that input x belongs to the set A with a high degree, while small membership values indicate that set A does not suit input x very well.

The form of membership function is dependent on the structure of the corresponding fuzzy set. Some known forms are triangular, trapezoidal, Gaussian, or bell-shaped.

Exhibit 2. General Fuzzy Expert System

General Fuzzy Expert System

Like classical sets, fuzzy sets maintain the operators of intersection, union and complement. These three operators are the most important and widely used and are analogous to the operators AND, OR, and NOT of classical logic. The function min is used to define the operator AND: the statement A AND B, where A and B are limited to the range (0,1), is defined as

img

where C is a fuzzy set that represents the union. As a similar manner the operation OR can be replaced using the max function, and the operation NOT A becomes equivalent to 1-A.

In its traditional rule-based formulation, an expert system is represented by a sequence of rules that describe the behavior of a natural system. Fuzzy sets and fuzzy operators are the subjects and verbs of fuzzy logic. These if-then rule statements are used to formulate the conditional statements that comprise fuzzy logic.

A single fuzzy if-then rule assumes the form:

if x is A then y is B,

where A and B are linguistic values defined by fuzzy sets on the ranges X and Y, respectively.

Expert rules use the classical logical operators AND, OR, and NOT connecting the linguistic variables. A set of different such rules constitutes an expert rule bank that behaves as the expert knowledge used to describe the operations and the control procedures of the system. In applications, most systems are considerably more complex than a simple list of rules and the decision-making process is not a simple application of these rules.

A fuzzy inference system employing fuzzy if-then rules can model the qualitative aspects of human knowledge and reasoning processes without employing precise quantitative analyses. Fuzzy inference is the process of formulating the mapping from a given input to an output using fuzzy logic. The mapping then provides a basis from which decisions can be made, or patterns discerned. The process of fuzzy inference involves membership functions, fuzzy logic operators, and if-then rules.

The general scheme of fuzzy expert systems consists of four parts: the fuzzification module, the defuzzification module, the inference engine, and the knowledge base:

The fuzzification module is associated with the transfer of the input signal form the crisp to the fuzzy representation world, while the defuzzification procedure converts the distributed fuzzy logic values into a single point solution value, which will constitute the output of the fuzzy expert system. The knowledge base part comprises the information given by the process operator in form of linguistic control rules, and finally an inference system makes inference by means of reasoning methods.

The most commonly used fuzzy inference systems are the Mamdani Model (Mamdani 1975) and the Sugeno Model (Sugeno 1988). In the Mamdani Model each fuzzy rule is:

If x is Ai the y is Bi

expresses a fuzzy relation Ci that is represented as a fuzzy intersection of the fuzzy sets Ai and Bi.

The Mamdani Model suggests that the contribution of a set of m fuzzy rules has the result of constructing a fuzzy relation C accomplished via the union of the individual fuzzy relations Ci.

The Sugeno fuzzy model develops its original formulation a systematic approach to generating fuzzy rules form a given data set. Typically, a fuzzy rule in this model has the following form:

If x1 is A1 and … and xN is AN then y = f(x1,…,xN)

where Ak, k = 1,…,N represent the fuzzy antecedent labels (fuzzy input sets). The big difference in this model is the functional type of consequence instead of the fuzzy consequence used in the Mamdani model.

Exhibit 3. Codification of Variables

Codification of Variables

System Approach

ESNEF presents several differences with the previously defined tools:

• It is designed specially for graduate and postgraduate education.

• User negotiate against computer.

• Web-based interface, client/server application.

• It introduces external factors as luck, changes in objectives, randomness, and so on.

• System considers personal relations or negotiation style of users. That is a main factor, as we know human factor is essential for negotiation.

• Inference motor is based in fuzzy rules instead of decision trees or mathematical formulation.

• Although maintaining the spirit of win-win solutions, it goes beyond, prizing efficient solutions, with no value left on the table.

Internal Model

Fuzzy rules present the advantage of its better fit to natural language. The composition of several variables produces nearly infinite different cases and it avoids the learning effect of the user to the behavior of the system. The main aspect to consider when introducing fuzzy in a model is the representation of the variables and the creation of the membership functions.

In this case main aspects to consider are:

• Offer. Final cost will be the main factor for every part. It is obviously the main confrontation interest as it is not possible a win-win strategy only for this factor.

• Time to agreement. To achieve an agreement early will be one of the objectives as more time implies generally more cost. Anyway the importance for every role is different. Administration is generally indifferent to time, owner could be interested in delay the negotiation to force a better price and Promoter will be more interested in a rapid conclusion as its project will be affected.

• Ways of negotiation. Process will be facilitated if relations between stakeholders are correct. System validates as excellent, cordial, normal, bad, and very bad relationships dynamically during the process according to the type of decisions that the partners are producing.

• Destiny. There are other factors generated randomly during the process to simulate noncontrolled factors that also affect a negotiation process in the project. That is for example the existence of formal errors during the process, changes in legislation or in the project, and so on. All those circumstances are introduced in a single fuzzy variable mixed together.

• Use of information. Internally the system knows the information available to the user and if he has accessed and used that information adequately. This information gives an idea of the analysis capability of the user and it is introduced for the generation of next sceneries.

Exhibit 4. Example of Window of the Application

Example of Window of the Application

System considers also time dependant variables including last offers from user and from the computer to consider efforts done during negotiation.

Every variable has been codified in functions, generally in five different levels, except the two first variables, which have more intervals due to their specific importance. With the functions and intervals, rules and operations where designed to create the system.

Negotiation System

There are many aspects to consider when facing a negotiation. After being identified, user must choose the role he will acquire:

• Proprietary: Person who will be expropriated.

• Public administration: Neutral organization in charge of the process.

• Promoter: Responsible of the project and cause of the expropriation.

Then a problem is selected and detailed information of objectives, difficulty, steps, and times is provided to the player. The main objective is always to find the best price for every problem, minimal in the case of promoter and maximal in the case of proprietary. System will act in an intelligent manner, offering first the simplest problems and avoiding the repetition of roles for the same user.

User has continuous access to tools remind him the objectives and the problem to solve, facilitating access to information, and reporting played time. User can interrupt the process any time and change to another case or to recover an old case. Also a virtual calendar is offered, as it is important to know times for allegations, actions, and so on. System is enriched with all legal aspects related to the expropriation process, as it is another field of interest for civil engineering training. For example it is important to discover errors in the expropriation process if acting as proprietary or to follow strictly the steps if acting as administration.

Negotiation process must be something more than a simple exchange of offers. Each party to a negotiation tries to influence the outcome in such a way as to maximize its own satisfaction, even if it depends of other party satisfactions. User finds a compromise or agreement when all negotiators jointly agree the combination of options across all issues after exchanging a sequence of offers. Obviously it requires giving up partly on some issues so as to gain on other issues. When user decides an action, system generate composed variables (as negotiation style) and introduce them in the inference motor producing a response to that action. Tackling issues one at a time can easily result in an impasse with on or more difficult unresolved issues. The holistic approach selected in ESNEF make possible to deal with all issues at once. System will also compares action with the optimal one and generates an internal index to qualify the negotiation capability of the player.

Exhibit 5. Results of Group Use Comparison

Results of Group Use Comparison

Those conditions produce a nonlinear structure of the problem, more complex to be learnt by the player and then opener and more creative. System could accept offer, propose a new offer, or even to break negotiation. Player must avoid arriving to an endless point due to immovable position or excessive tension in negotiation. Moreover player must try to agree in a position positive for him, as it will be ranked with a percentage that if under 50 percent means a worse agreement than initially.

In order to provide an advice, the figure of an expropriation court has been also created to inform of the most equitable price at that time.

Simulation finish when there is not possibility of agreement because one of the parts has denied or if system determines lack of interest, indicating time played, final price, and some indicators of the goodness of the actions undertaken.

Implementation of the System

ESNEF system is envisioned to be either a stand-alone system for research and advanced training programs or in combination with other tools as part of a platform for decision-making and negotiations.

The application is designed in a client/server format in order to be executed remotely from any point in a connected network. Server is based on Apache httpd (http://www.apache.com) running on LINUX operating system. That selection reduces the cost as both tools (operating system and server) are freely distributed under GNU license.

Data was managed by a My SQL database in the first versions, due to its free distribution again, but currently it is prepared for its use with Oracle RDBMS in order to make system more adaptable. Database is the “heart” of the simulator and contains all the necessary information to make the system works. Queries are programmed with PHP technology embedded in the HTML pages, also freely distributed. Cookies are used to avoid the access to any part of the game without registration and to remember previous behavior of users. Every new user must register and with form/multipart technology system generate a unique alphanumeric code.

Clients work on a standard navigator (Microsoft® Explorer®, Netscape®, and so on) without needing any installation on it. It reduces drastically the availability and time to work of the system, but especially it allows continuous adaptation without maintenance costs and full update of every client.

Application

System has been used for training in project management to different levels. Here we present a comparison of two different groups: twenty professionals in project management from a big European company (more than 30,000 employees), all them with more than five years experience in project development as team responsible and seventy-three civil engineering students in the last year of their careers. Age average is almost double (forty-one) for engineers. Students had not work experience, but they have received specific formation on project management. Both groups used the program until getting an expected value superior to the first offer (ranked as 50 percent). Exhibit 5 shows the results of the tests and the opinion of the users.

In general it is easy to observe that experience is important to solve the problem. Engineers must repeat the simulation only two times before they find a solution and their solutions are more optimal (78 percent compared to the maximum price available). Also their strategy, internal parameter determined by the game, is better (89 percent) as they proceed considering more aspects than immediate information.

Opinion of the users is resumed in two factors: global valuation of the activity and interest in repeating a similar activity in other formative actions. In both cases opinion is considered positive. Values are more positive in the case of engineers as they realize of the importance of negotiation and they are more concienciated. The activity for students is part of their works to finish their careers although his opinion is generally positive and they welcome more interactivity in teaching.

Conclusions

This paper has two levels of research interest: the use of game’s theory for learning of an important part of project management as negotiation in a client-server approach and the introduction of new techniques as fuzzy logic and knowledge-based systems to create new environments and to evaluate the behavior of the student in order to evaluate his progress and reasoning.

The system presented here is interesting for trainers in project management as the importance of distance or computerized learning is growing every day and it will impose in a close future. But also it is interesting for companies as it is a technique to use for the formation of their own teams.

From the second point of view, the use of artificial intelligence technique makes possible to build simulation tools capable to imitate the real project environment and it can be used for distance project management learning. Fuzzy logic reveals as a very adequate techniques due to its capability to generate models similarly to human behavior.

This kind of teaching tool is well understood and accepted by the users, independently if they are experienced or still undergraduate, although results shows that experience helps to afford the problem with more warranties.

System will contribute to the development of training in the field of project management in an aspect as relevant as negotiation, but the innovative way it is presented will produce an added interest to the users for the project management methodologies.

Research is not finished and further development is needed to introduce more variables in the fuzzy rules and to optimize solutions with more refined non-linear methods as evolutionary strategies.

References

Cano, J. L., and M. J. Saenz. (1999). Development of a project simulation game. Project Management Journal 5 (1): 37–41.

Jang, J. S. R., and C. T. Sun. (1995). Neuro-fuzzy modeling and control. Proceeding of the IEEE 83 (3): 378–406.

Jonassen, D. H. (1991, September). Evaluating constructivistic learning. Educational Technology: 28–33.

Mamdani. (1975). Applications of fuzzy logic to approximate reasoning using linguistic synthesis. IEEE Trans. Computers 26 (12).

Poundstone, William. (1993). The Prisoner’s Dilemma. New York: Doubleday.

Simbandumwe, Jean-Paul. Tools for Developing Interactive Academic Web Courses. Accessed at http://www.umanitoba.ca/ip/tools/courseware.

Sugeno, M., and G. T. Kang.(1988). Structure identification of fuzzy model. Fuzzy Sets and Systems 28: 15–33.

Young, M. F. (1993). Instructional design for situated learning. Educational Technology Research & Development 41 (1): 43–58.

Zadeh, L. A. (1965). Fuzzy sets. Information and Control 8: 338–353.

This material has been reproduced with the permission of the copyright owner. Unauthorized reproduction of this material is strictly prohibited. For permission to reproduce this material, please contact PMI or any listed author.

Proceedings of PMI Research Conference 2002

Advertisement

Advertisement

Related Content

Advertisement

Publishing or acceptance of an advertisement is neither a guarantee nor endorsement of the advertiser's product or service. View advertising policy.