What is decision automation?
by Daniel J. Power
Decision automation means software is used to make decisions. The concept is deceptively simple and intriguingly complex. On the surface the idea is to write a computer program that uses rules and criteria to make decisions. A decision automation system replaces and eliminates the need for a human decision maker in a specific decision situation. Business rules and programmed instructions are triggered by inputs and events and then the program "makes" contingent choices. The greatly expanded and evolving computing infrastructure makes it increasingly cost effective to apply decision automation in situations where that had been prohibitively costly. A decision automation system is NOT a decision support system.
Davenport and Harris (2005) claimed "After decades of anticipation, the promise of automated decision-making systems is finally becoming a reality in a variety of industries." They noted that bank credit decisions can be automated. They also noted the increased used of automated decision making for home loan lending decisions by DeepGreen Financial. Sadly as noted in Power (2014) automated decision making may have contributed very directly to the financial subprime mortgage crisis.
In my Decision Support Systems concepts book (Power, 2002), I argue “routine decision situations should be analyzed and 'programmed' as much as is possible and they should be supported in most situations by technology. The potential rewards from improving routine, recurring decisions are usually very large (p. 39).” Lending decisions are perhaps too risky to be considered routine. Just because a decision is made repeatly and criteria are applied in a standard way doesn't mean the decisions then becomes routine and low risk.
From a narrow perspective a decision is a choice among defined alternative courses of action. Hence decision automation systems make choices among predefined alternatives. From a broader perspective, a decision involves the complete process of gathering and evaluating information about a situation, identifying a need for a decision, identifying or in other ways defining relevant alternative courses of action, choosing the “best”, the “most appropriate” or the “optimum” action, and then applying the solution and choice in the situation. Hence decision automation systems can also help automate all or part of a specific decision process. Automation refers to using technologies including computer processing to make decisions and implement programmed decision processes. Typically decision automation is considered most appropriate for well-structured, clearly defined, routine or programmed (cf., Simon, 1960) decision situations.
An automated decision process occurs without any “human” intervention. Decision making procedures are programmed and then appropriate programs evaluate stored or “real time” data from sensors. An algorithm based on quantitative, logical, heuristic, statistical, and/or artificial intelligence technologies processes the data. The algorithm specifies one or more actions to apply in the situation. The actions are taken by effectors including human or machine actors. An effector may change the value in a database, send a message or an alert, move an object or play a message.
Human decision makers determine the alternatives, rules, models and methods used for making choices and completing decision tasks in programmed decision situations. Conceivably decision automation programs can “learn” from successes and failures and automatically improve and update the relevant stored procedures, rules or likelihoods.
Overall decision automation is a set of concepts, a related set of technologies, a set of methods and design tools, and an ambitious, general “goal”. The range of decision tasks that can be automated has increased as technologies and design tools have improved. This technology evolution has also raised aspiration levels and created more challenging development goals. The overriding goal of a decision automation project is to replace human decision makers in programmable decision situations where it can be demonstrated that the computer program's decision is at least as good as that of all or most human decision makers. The working assumption is that decision automation will be cost effective when compared to an equally skilled human decision maker in a programmed decision situation.
Decision Automation Application Areas
The major improvements in computer processor speed, faster and larger capacity storage technologies, improved sensors, the ubiquity of the Internet and World-Wide Web and increased reliability of information technologies has greatly increased the number and diversity of decision automation applications and that trend is accelerating. Decision automation can assist in an air traffic control environment, in flexible manufacturing systems, in petroleum refining, in high-speed sorting systems, in tax decision automation, in intelligent monitoring and decision making in intensive care situations, in fruit grading, in real-time notification, in credit approval automation, in airborne collision avoidance systems, in building automation and facility management systems, in hardwood log breakdown, in laboratory management and automation systems, and in many other well understood decision situations.
Plant Automation and Decision Support is important in flow and process industries. Rob Spiegel, Contributing Editor to AutomationWorld.COM, writes “Web technologies are proliferating through plant automation systems, letting managers review production and control data from anywhere they can access the Internet. So far, the bulk of Web-based plant monitoring is used to obtain production data (June 2004, page 34).”
So what are the major research and content topics associated with decision automation? The fundamental research topic is developing algorithms for machine processing, the development of the “brain” of a specific decision automation system. Algorithms especially need to exploit parallel processing. Development technologies that are used for automating decision processing include AI systems, Bayesian networks, intelligent software agents, and neural networks. The storage or “memory” of the system is important, but major improvements have temporarily reduced the need for new breakthroughs. The goal for enhancing storage is always higher capacity in smaller devices with faster read/write capabilities. Sensors, the “input” devices of decision automation, are an important research area. We need better visual systems and more automated data collection. Radio frequency identification (RFID), global positioning systems (GPS), and environment monitoring technologies are all contributing to the expansion of sensing technologies. Finally, decision automation requires a means to communicate "the machine's" decision. This task is the realm of “effectors”, the “output” devices that can modify a database, control a robot arm, or stop a production line.
Bucklin, Lehmann, and Little (1998) predict that in the coming decades, a growing proportion of marketing decisions will be automated. In 2001, Henry Morris, IDC Vice President, wrote “Only when decision support moves to decision process automation, can we improve quality of decisions.” Randy Fields in a DSSResources.COM Thought Leader Interview expressed a similar, but more sophisticated view of the need for decision automation. Randy noted “What we have done is to take control away and let machines make the decisions. It is a fundamentally different view of the role of technology. We haven't built thermostatic systems that flash to people, little engineers running around, and tell them in better real time or audibly what the temperature in the room is. We have the sensors, coupled with processors make decisions and implement them. The humans just set the strategy.”
Decision automation is important and it has an expanding role in organizations, but there still remains a major role for human decision makers and hence for decision support systems. Decision support can often improve decision quality in semi-structured situations where decision automation is not feasible or is undesirable.
To support the growth of this knowledge area, I created a website called Decision Automation Resources (DecisionAutomation.COM) in 2004. This site focuses on using quantitative models, heuristics, statistical approaches and Artificial Intelligence technologies to make decisions in highly structured and routine decision situations. DSSResources.COM and PlanningSkills.com remain the “flagship” websites for people interested in improving decision making and planning, but now developers and students can also visit DecisionAutomation.COM.
DecisionAutomation.COM is still evolving. If you have any suggestions for improving the site, please email me at email@example.com .
Bucklin, R.E., D.R. Lehmann, and J.D.C. Little, “Decision Support to Decision Automation: A 2020 Vision”, MSI working paper, 1998.
Davenport, T.H. and J.G, Harris, "Automated Decision Making Comes of Age," MIT Sloan Management Review, Summer 2005, July 15, 2005 at URL http://sloanreview.mit.edu/article/automated-decision-making-comes-of-age/
Morris, H., “Decision Support vs. Decision Process Automation,” IDC Viewpoint, August 2001, http://www.idc.com/getdoc.jsp?containerId=VWP000060
Power, D. J., Decision Support Systems: Concepts and Resources for Managers, Westport, CT: Greenwood/Quorum, 2002.
Power, D. J., “Randy Fields Interview: Automating 'Administrivia' D ecisions”, DSSResources.COM, 04/09/2004.
Power, D., Did decision automation cause the subprime mortgage crisis? Decision Support News, Vol. 15, No. 7, March 30, 2014 at URL dssresources.com/faq/index.php?action=artikel&id=299.
Power, D. J. and R. Sharda, Chapter 87 Decision Support Systems, in Nof. S. (ed.) Handbook of Automation, Berlin: Springer, 2009, pp. 1539-1548.
Simon, H.A. (1960). The New Science of Management Decision. New York, NY: Harper and Row.
The above response is modified and updated from Power, D., What is decision automation? DSS News, Vol. 5, No. 14, July 4, 2004 (August 5, 2011 and July 15, 2014).
Last update: 2014-07-18 05:15
Author: Daniel Power
You cannot comment on this entry