The Blackwell Encyclopedic Dictionary ofsoonang.com/wp-content/uploads/2011/04/1996-Ang-AG… ·  · 2011-04-22Encyclopedic Dictionary of Management Information System.s Edited by

THE BLACKWELl. ENCYCI.OPEDIA OF MANAGEMENT

About the Editors

Cary L. Cooper is Professor of Organizational Psychology at the Manchester School ofManagement (UMIST), UK. He has also been appointed Pro-V ice-Chancellor at the Universityof Manchester Institute of Science and Technology (UMIST). He is the author of over 80 books, haswritten over 250 scholarly articles and is editor of The Journal of Organizational Behavior. He is alsothe Founding President of the British Academy of Management.

Chris Argyris is the James B. Conant Professor of Education and Organizational Behavior at theGraduate School of Business, Harvard University. He has written many books and receivednumerous awards, including the Irwin Award by the Academy of Management for lifetimecontributions to the disciplines of management. Recently, the Chris Argyris Chair in SocialPsychology of Organizations has been established at Yale University.

About the Volume Editor

Gordon B. Davis is the Honeywell Professor of Management Information Systems at the CarlsonSchool of Management, University of Minnesota. He is recognized internationally as one of theprincipal founders and intellectual architects of the academic field of Management InformationSystems. He has authored 21 books and over 150 scholarly articles. He has been involved inestablishing academic journals, scholarly conferences, and model curricula for the field of informationsystems.

~~00'-

The BlackwellEncyclopedic Dictionary ofManagement Information

System.s

Edited by Gordon B. Davis

Carlson School of Management

13BLACKWELLBusiness

T·f AGENCY THEORY APPLIED TO INFORMATION SYSTEMS 3

make decisions regarding corrective actions thatmay be needed.

WII.l.I.\~1 D. N.\NCE

ACM The Association for ComputingMachinery is the largest, broad-based interna-tional computer and information system society(Ul!' ASSOCIATIO:--lSA:--It>S(K:IETIES FOR I:--IFOR·MATION SYSTEMS I'ROFESSIO:'-JAI.S).

ADA A general purpose programming lan-guagesponsored by the United States Depart-ment of Defense. It is especially suited for (heprogramming of large, long-lived systems with aneed for ongoing maintenance. It supportsmodern programming structured techniquesand concurrent processing.

agency theory applied to informationsystems Agency theory examines the con-tracts between a party (the principal) whodelegates work to another (the agent). Agencyrelations become problematic when the principaland agent have conflicting goals and when it isdifficult or costly for the principal to monitor theperformance of the agent. When goals areincongruent, the agent is assumed to have adifferent set of incentive structures from theprincipal; the agent will consume r-:rquisites OUI

of the principal's resources and make suboptimaldecisions. These activities produce efficiencylosses to the principal. To counter these losses,the principal designs contracts to align the goalsat the lowest possible costs. Costs can arise fromproviding incentives and from monitoring toensure that the agent is acting for the principal'sinterests.

Agency theory can offer insights for informa-tion systems. First, principals can designinformation systems to monitor the actions ofagents. Electronic communication systems,electronic feedback systems, and electronicmonitoring systems are examples of monitoringdevices that can be implemented to ensure thatagents' behaviour is aligned with principals'interests.

Secondly, information systems professionalsthemselves often enter into agency relationshipswith other stakeholders in organizations andagency problems can arise. Important examplesof such agency relationships include systemsdevelopment, outsourcing, and end-user com-puting.

Syslems Derelopmcnt

As principals, users often engage informarionsystem (IS) professionals as agents to developinformation systems on their behalf. Due 10 alack of understanding and knowledge of eachother's domain, goal conflict may arise betweenthe two parties. To reduce agency costs, one or'both parties must try to narrow goal differences.IS professionals can invite users to participatemore actively throughout the development life-cycle. This gives the users more opportunitiesto verify requirements and ensure that the finalsystem is aligned with user needs. Further,users may request that the information systemproduce information-rich documentation so thatmonitoring is made easier and more readilyavailable to users.

Dutsourang

In any outsourcing arrangement, the clientcompany (principal) is usually motivated toshift its IS operations to external vendors whocan carry out the work at the lowest possiblecost. The vendor, on the other hand, may belooking for high profit in the arrangement.There is thus an economic goal conflict. Toprotect its interests, the client will increase itsmonitoring of the vendor. This can be achievedby requesting regular operational performancemeasures from the vendor, frequent meetingswith the vendor to review progress of out-standing projects, and independent auditors toreview benchmarks and internal control, of thevendor.

End-user Computing.

Agency theory can help explain the dynamics ofend-user computing. End users develop infor-mation systems themselves with little ISinvolvement. End-user computing, interpretedin agency theoretic terms, is a mechanism forreducing agency problems by eliminating the

4 AIS

agency relationship between the user and ISprofessional.

SOON ANG

AIS The Association for Information Systemsis an international society for informationsystem academies (sa ASSOCIATIONS ANDSOCIETIES FOR I:--IFORMATIONSYSTEMS PRO-FESSIO:-;.\LS).

artificial intelligence The attempt to pro-gram computers to perform tasks that requireintelligence when performed by humans isknown as artificial intelligenc«. Examples ofsuch tasks are visual perception, understandingnatural language, game-playing, theorem-prov-ing, medical diagnosis, and engineering design.

Beginning in the late 19505, AI researchershave modeled a variety of problems (such asplaying checkers or proving theorems inmathematics) in terms of state space search. Astate denotes a particular configuration of thecomponents of a problem. The position ofpieces on a chess board and the structure ofterms in a mathematical expression are examplesof states (for the problems of chess-playing andtheorem-proving, respectively). The applicationof a permissible operator (such as a legal movein the game of chess or an expansion of terms ina mathematical expression) alters the state of aproblem. The set of all possible states, togetherwith the operators that enable transitions amongthem, constitutes the state space representationof a problem.

The solution of an AI problem consists of asearch through the state space, i.e. the succes-sive application of operators until the final stateof the problem matches the desired goal state (acheckmate in chess or the simplest expression ofa theorem). Unless the problem is very limitedin scope (e.g. playing ric-rac-toe), the state spaceis hopelessly large for an exhaustive search (thegame of chess has more than 10110 states).Additional knowledge (beyond the rules of thegarne) is required to guide the state space searchin promising directions. This search controlknowledge is commonly called heuristic knowl-edg«. The process of problem-solving in AIdescribed above is called heuristic search (Newell& Simon, 1976).

Chess-playing and theorem-proving arexamples of tasks where the careful appbcatioiof logic has to be supplemented by heuristi,knowledge to produce an efficient solution. .\!AI research progressed, it was discovered thaispecialized tasks, such as diagnosis, design. andplanning, require even more knowledge toformulate (in state space terms) and solve(through heuristic search). In order for 1.:00 ••.1-edge to facilitate the solution of an otherwiseintractable problem, the knowledge must berep-esented in a suitable form for use by acomputer program. Methods of reasoning aboutthe knowledge to apply it to a particular siruarionmust also be specified. The representation ofdomain knowledge and efficient methods ofreasoning with it IKve become central concernsof AI since the 19705 (Feigenbaum & McCor-duck, 1983). Certain formalisms, including if-then rules, semantic networks, frames, andpredicate logic, have been developed to repre-sent and utilize knowledge efficiently in prob-lem-solving.

AI methods have been successfully applied toproblems in computer vision, robotics, kno .••.l-edge-based systems (sa EXPERT SYSTUtS;KNOWLEDGE BASE), understanding naturallanguage and machine leaming (the extractionof patterns from large volumes of data) .. -\.1-based computer systems have been successfullydeployed in manufacturing to support thedesign and diagnosis of products and processes.In services, AI has been applied to a variety oftasks, including medical diagnosis, fmancialstatement analysis, and logistics management.In addition to dedicated AI systems, .\1techniques have also been used to improve theuser interfaces of conventional infonnationsystems.

Sa alsa Cognitive science and informationsystems

Bibliography

Newell, A. & Simon, H. A. (1976). Computer scienceas empirical inquiry: symbols and search. c"m=-nications of th« ACM, 19 (3), 113-26.

Feigenbaum, E. A. & McCordud, P. (1983). To<Fifth Generation: Artificial J"ldl;gcna and Jap"",