Decision Support Systems

A decision support system (DSS) is a computer-based arrangement that helps a person or a group make a decision they could not make as well unaided.

8 frameworks, from 1960 to today · Information Systems · Business, Management & Finance

still a working view ended · each leads to its entry below

Overview

A decision support system (DSS) is a computer-based arrangement that helps a person or a group make a decision they could not make as well unaided1. The term covers a wide range: a spreadsheet model a manager builds to test a budget, a hospital alert that flags a dangerous drug interaction, a room of executives voting anonymously through keypads, a dashboard that predicts which customers will leave. What unites them is the design goal: not to automate the decision, but to leave judgment with the human while supplying computation, data, or structured process that the human could not supply alone. The field matters because so much organizational computing fails at exactly this point—systems that produce numbers nobody uses, or that quietly displace the judgment they were meant to support.

Bounded rationality and the founding idea

The intellectual root is Herbert Simon's work on administrative decision-making in the 1940s and 1950s. Simon argued that real decision-makers are boundedly rational: they cannot survey all alternatives or compute all consequences, so they satisfice—accept a good-enough option rather than search for the best. If human rationality is limited, then tools that extend it have a clear job. Simon's 1960 book The New Science of Management Decision distinguished programmed decisions (routine, rule-governed) from non-programmed ones (novel, unstructured), and framed the computer as a way to move more decisions toward the programmed end.

The phrase "decision support system" itself emerged in the early 1970s, notably in work by Gorry and Scott Morton, who combined Simon's programmed/non-programmed distinction with Anthony's categories of managerial activity (operational control, management control, strategic planning). Their two-dimensional framework located DSS in the semi-structured middle: decisions where some part can be formalized and computed but the whole cannot. This semi-structured framing became the field's defining charter. A fully structured problem could be automated; a fully unstructured one was left to judgment; DSS carved out the space in between, where a model could inform a human decision-maker without replacing them.

Model-driven DSS and the component architecture

The first generation of systems, from the 1970s into the 1980s, was model-driven: the core of the system was a quantitative model—financial planning, optimization, simulation—that the decision-maker could manipulate. The canonical example is the spreadsheet, whose business adoption followed the 1979 release of VisiCalc and later Lotus 1-2-3; it let a manager change an assumption and see the consequences immediately, which is model-driven decision support in its purest everyday form. More elaborate systems embedded linear programming, risk analysis, or project models behind a friendly interface.

Designers of this era converged on a component architecture that is still the field's basic vocabulary. A DSS was said to consist of a data management component (a database and its query facilities), a model management component (a library of models plus the means to build, run, and link them), a dialog or user-interface component (the language through which the user interacts), and often a knowledge component. The influential formulation came from Sprague and Carlson's Building Effective Decision Support Systems (1982), which also stressed that DSS should be built iteratively and adaptively rather than specified once—a contrast with the waterfall conventions of mainstream transaction-system development. The architecture's lasting contribution was conceptual: it told builders that the model, the data, and the conversation with the user were separable concerns, each with its own design problems.

Group decision support and the meeting room

By the late 1980s attention shifted from the lone manager to groups1. Group decision support systems (GDSS) combined computing, communication, and decision techniques to improve meetings—especially meetings plagued by dominance, anchoring on the first proposal, and pressure to conform. The best-known research program was at the University of Arizona, where Jay Nunamaker and colleagues built "decision rooms" in which participants at networked terminals could brainstorm anonymously, comment on proposals, and vote, with results projected on a shared screen. The anonymity was the point: it let junior staff criticize a senior manager's idea without reprisal, and it let ideas be judged on merit. Field studies of these rooms reported gains in the number and quality of ideas and in meeting satisfaction, though the effects depended heavily on facilitation and task. The line of work fed directly into computer-supported cooperative work and into today's collaborative platforms, and its core insight—that the social process of deciding can be engineered as deliberately as the analytical content—remains a live design principle.

Knowledge-based and expert systems

A parallel strand drew on artificial intelligence. Expert systems of the late 1970s and 1980s, such as MYCIN for diagnosing infections and XCON for configuring computer orders, encoded human specialists' rules as if–then production rules and could reason through a problem, explain their reasoning, and ask for missing facts. For DSS researchers the appeal was the knowledge component: a system that carried domain expertise could support decisions in areas too complex for a numeric model. The resulting knowledge-based DSS combined an inference engine, a knowledge base, and an explanation facility with the conventional data and dialog components.

The approach's limits became as instructive as its successes. Eliciting expert knowledge proved slow and expensive; rule bases became brittle at their edges and hard to maintain; and systems that performed well in narrow domains did not generalize. The rise of machine learning from the 1990s onward shifted the emphasis from hand-coded rules to statistical models learned from data—but the goal of embedding expertise in a decision aid, and the demand that such a system be able to explain itself, carried straight over.

Data-driven DSS and business intelligence

The other major branch grew from data rather than models. As transaction systems accumulated years of records, a new question arose: how to let managers explore what had happened and why. Data-driven DSS—query-and-report tools, online analytical processing (OLAP, which lets users slice aggregated data by dimension such as product, region, and month), and executive information systems with their drill-down screens—answered it. The supporting infrastructure was the data warehouse, a consolidated store built specifically for analysis rather than transaction processing, and the practice of extracting, transforming, and loading data into it2.

Through the 1990s this cluster was rebranded business intelligence (BI): a stack running from warehouse and ETL through reporting and dashboards to scorecards tracking key performance indicators. BI's promise was organizational visibility—a single version of the truth about performance—and its recurring failure was the same one the field had always faced: dashboards that were built but not used, metrics that were gamed, and reports that answered yesterday's question. The lesson that decision support is a sociotechnical problem, not merely a technical one, was learned repeatedly here.

Clinical decision support

Medicine became the field's most consequential application domain. Clinical decision support systems (CDSS) embed patient-specific knowledge into care: drug–drug interaction and allergy alerts, dosing calculators, reminders for preventive care, diagnostic aids, and order sets that guide prescribing. Early landmark systems included the HELP system at LDS Hospital in Salt Lake City, developed from the 1970s, and the Regenstrief Medical Record System, which from the 1970s onward delivered reminders to clinicians at the point of care. The 1990s brought evidence that computerised physician order entry with decision support reduced serious medication errors, and adoption spread with electronic health records.

Clinical DSS also produced the field's sharpest evidence about unintended effects. Alert fatigue—clinicians dismissing so many low-value warnings that they miss the important ones—became a documented hazard, and studies found that poorly designed alerts could increase rather than reduce risk. The domain forced the field to confront questions of liability, explainability, and workflow fit that now recur wherever decision support is deployed.

Business analytics and predictive decision support

From the 2000s, cheaper storage, faster computation, and abundant data moved the center of gravity toward prediction. Business analytics extended BI from describing the past to forecasting the future: churn models, credit scoring, demand forecasting, fraud detection, and, later, prescriptive analytics that recommend actions rather than merely predict outcomes. Machine learning displaced hand-built models in many applications, and the DSS label increasingly attached to the decision layer atop an analytics pipeline. The open questions are the field's oldest ones in new dress: how much to trust a model whose reasoning is opaque, how to keep the human genuinely in the loop rather than rubber-stamping a recommendation, and how to design systems that improve judgment instead of eroding it. The component architecture, the semi-structured charter, and the group-process insight of the 1970s and 1980s remain the vocabulary in which those questions are still posed.

References

  1. Decision support system (introduction, History), Wikipedia
  2. Data warehouse, Wikipedia

The frameworks of Decision Support Systems

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1960-Present
  2. 1975-Present

    Clinical decision support system

    How can computers help clinicians avoid errors?

    Fly to its tower in Nousfarer

  3. 1975-1995

    Model-Driven DSS

    How can a model inform a manager's choice?

    Fly to its tower in Nousfarer

  4. 1978-1995

    Expert system

    How can a machine reason like an expert?

    Fly to its tower in Nousfarer

  5. 1980-Present
  6. 1985-2005
  7. 1990-Present

    Business intelligence

    How can organizations see their own performance?

    Fly to its tower in Nousfarer

  8. 2005-Present

    Business analytics

    How can data predict what will happen next?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Bounded Rationality and Semi-Structur…Clinical decision support systemModel-Driven DSSExpert systemDSS Component ArchitectureGroup Decision Support SystemsBusiness intelligenceBusiness analytics

Beyond the field

Where Decision Support Systems's ideas came from in other fields, and what they led to there: the strongest 12 of 30 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Decision Support Systems is a planet in Nousfarer

Its city has a tower for each of these 8 frameworks. Land, walk in, and climb one concept at a time.

Land on Decision Support Systems

Related fields

Fields its ideas are linked to

More in Information Systems

See where Decision Support Systems sits on the Map of Human Knowledge