Decision Automation
Decision support supplies information for a person to use; decision automation also selects or carries out an action. Automation can use fixed rules, statistical models, or learning systems. Repeated decisions with clear outcomes and feedback are useful candidates for evaluation.
THE IDEA
Decision Automation
Decision support tools help a person choose. Decision automation goes further by selecting or carrying out an action, such as routing an order or adjusting a production process. A system can perform both roles in different parts of a workflow.
Automation may use fixed business rules, probabilistic models, or machine learning. Using a rule does not imply that a system learns, and probabilistic reasoning need not be Bayesian. Learning systems require a way to compare their choices with later outcomes. Repetitive decisions with clear objectives and useful feedback are therefore easier to evaluate than novel decisions with ambiguous results.
The author places decision automation in a history of mechanization beginning with the steam engine in the late 18th century. Automating judgment raises questions beyond whether a machine can perform a manual task: can it recognize a situation, choose an appropriate response, and handle cases outside its design assumptions? Autonomous vehicles illustrate those questions because they must act on sensor information in changing environments.
The source also discusses quantum entanglement 5, quantum superposition 6, and unfamiliar quantum principles 7 as possible contributors to future technologies. These possibilities do not establish that automated systems will become safe or that public concerns will disappear.
Noise-cancelling headphones are a different case. They produce sound that interferes destructively with ambient sound. This is classical wave superposition, rather than an application of quantum computing, and it does not substantiate the source’s predictions about autonomous decisions.
When to use it
Automating quality control in manufacturing lines
Algorithmic trading in financial markets
Diagnosis and treatment recommendations in healthcare
Autonomous vehicle navigation
FROM THEORY TO EVERYDAY LIFE
In practice
Autonomous Vehicles
An autonomous vehicle uses sensor information to choose actions such as braking or steering. Evaluating it requires evidence about its performance, failure modes, and responses to unfamiliar conditions.
Noise-Cancelling Headphones
Noise-cancelling headphones generate sound that interferes destructively with ambient sound. The example illustrates classical acoustic superposition, not quantum superposition or quantum decision automation.
A MOMENT TO REFLECT
Put it to work.
What factors determine whether a decision is suitable for automation?
What evidence would distinguish a demonstrated improvement in decision automation from a speculative claim about a new technology?
Why might society be more critical of machine errors than human errors in high-stakes decisions?
Often confused with
CONTINUE THE CONVERSATION
Source texts
- Evidence-Based Decision-Making How to Leverage Available Data and Avoid Cognitive Biases (Andrew D. Banasiewicz)