Information vs. Decision Making

Information and decision-making are intimately connected and interdependent.

At today’s digital dynamic, information is abundant and even explosive, the business has become over-complex also hyper-connected, what’s the correlation between information and decision making. If data-decision pair works fine to advocate analytics based culture; as information is the processed data; and knowledge is processed information, how about the information-decision pair? It's not necessarily information, but the business insight leads to the right decision making. So how to measure information accuracy as well as decision-making effectiveness?

Decisions are based on information and generate information. The amount of data required for a decision, and the amount of information generated by a decision, can both be measured in bits. So a “yes/no” decision generates one bit of information, a decision consisting of an integer between 1 and 1024 generates 10 bits. The amount of information contained in a message was determined by its “surprising-ness”: that is, the number of different messages which might have occurred. The measure is logarithmic. In addition, the words like "information" and "data" have a variety of meanings: colloquial, professional and technical, "information" as input to the decision-making does not absolutely determine the decision but allows the decision-maker to exercise their judgment.

Information and decision-making are intimately connected and interdependent. There is more to it - information, principally more than data and knowledge alike, carries fundamentally higher entropy than the other two do. The effective information must come from somewhere; it is of course provided by the decision-maker, in the knowledge or expertise which they bring to bear on the problem. Or to put it another way, for decision-making to be effective, the decision-maker must have enough knowledge to make their decisions rich in information and significantly different from the available data.

Decision makers as information agent: If you look at the information life-cycle, any decision-makers as information agents are key in converting/processing data and knowledge into information as well as consuming that information in their decision-making. And no information can ever be generated without agents. Information has something to do with (products of) reifications and ramifications by individual agents. And the critical phase when information is created is when agents/decision-makers frame problems or more exactly when they DISCOVER problem structures. All the rest of the time, they generate only either new/redundant data and/or knowledge updates.

Information is situated between data and knowledge. The interesting metric is the relative entropy between the input data and the decision result (otherwise known as the Kullback–Leibler divergence). Obviously, if a decision-maker just passes a message on without influencing it in any way, the effective information provided is zero, regardless of the amount of information in the message. The data and knowledge are being convergent in terms of what action ensues (for example in analytics), whereas information is divergent as the right decision-making process essentially is. Useful decision-making occurs when the decision results are different from the input data, which is why effective information is measured as a divergence.

Data just ENDORSES action and knowledge just DETERMINES/SHAPES action, while it is information that does NOT affect action in any direct manner, instead it just CONDITIONS of action which makes it key to decision-making or choice, planning, and action (selection) ultimately. You would have to treat the business context as part of the grammar for the purposes of the analysis, but that makes sense: If data-decision pair is obviously correct (true) given the context, whereas an information-decision pair is not. If "data" is taken to mean input used in a decision which contributes no effective information to it. And yes, adding information to data is something done by decision-making agents.

Information applies to the context and environment in which decisions are made. Information, with the inclusiveness of data as input, is primary drivers of decisions when they apply to automated systems, not human beings. In the human context, information drives awareness, which can include all of these characteristics, uncertainty, surprise, difficulty, and entropy, although it can also trigger a sense of confidence, confirmation, validation, verification. Once you get to human beings, models become poor approximations for how they respond to inputs of all sorts, although they are still useful in a general sense. If you take the proposed choice/action as the subject and the data/information/knowledge as the predicate, then propositions with data and knowledge as predicates will be analytic while those with information will be synthetic. There is a theory of decision information to be developed here, perhaps based on the well-established information theory and computability. It is described in the AI literature as "non-monotonic reasoning" - where a decision made on the basis of incomplete evidence is revised as new evidence arrives. There is also a nastier type of surprise, where the decision-maker realizes that the knowledge s/he has is not adequate to make the decision (their decision logic is incomplete). This provides an opportunity to change decision-making behavior. These two situations are both above and beyond the concept of uncertainty in information theory. 

Information is the lifeblood of business, and decision makers are the information agent who can master information to capture not only just knowledge but business insight, well mixed with the right amount of guts, in order to make the right decisions at the right time to lead their business forward.