In recent years, artificial intelligence (AI) decision-making and autonomous systems became an integrated part of the economy, industry, and society. The evolving economy of the human-AI ecosystem raising concerns regarding the risks and values inherited in AI systems. My research on planning and decision making in oversubscribed systems investigates the dynamics of creation and exchange of values and points out some gaps in perception of cost-value, knowledge, space and time dimensions. In this blog, I would like to suggest a deeper look at the role of data in the decision-making process and to provide some of the observations from my recent paper on “Economics of Human-AI Ecosystem: Value Bias and Lost Utility in Multi-Dimensional Gaps” (2018).
From Data to Decision
The buzzword big-data is used many times in the context of the progress of technology, companies, and people collect data warehouse; the bigger the data, the merrier. The point in data is not just its existence, data is the key component for making decisions and decisions can be rational or irrational. Decision actions can be rational or irrational.
Following Marwala (2013, 2014, 2015) let us define a decision as rational if it resulted in a global optimum, based on logical principles and derived from complete relevant information. A decision that based on irrelevant or incomplete information is irrational, and can not guarantee guidance for the global optimum. The incompleteness of effective information results in an irrational decision (Marwala, 2013). Rational decision-making process consists of rational decision actions, and the entire process optimized in time and results in global utility optimum (Grüne-Yanoff 2012; Marwala 2014;2015).
The objective of developing rational decision-making systems comes with the rush for big data and the completeness of information for making decisions. But what if the data is too big? To answer this question let us have a deeper look at the role that data plays in the decision making process.
The relation of information and knowledge is modeled in the DIKW hierarchy (data — information — knowledge — wisdom) as illustrated in Figure 1.
Some researchers agree that the origins of DIKW are in T.S. Eliot’s poem The Rock (1934), in which he wondered;
Where is the Life we have lost in living?
Where is the wisdom we have lost in knowledge?
Where is the knowledge we have lost in information?
Years later, the DIKW hierarchy got attention in the literature of research (Zeleny, 1987; Cooley, 1980; Ackoff, 1989). An in-deep survey by Rowley (2007) shows several observations and extensions to the basic model. Ackoff (1989) made an example in practice to the theory by providing a refined, one-page paper explanation to the relations between the DIKW hierarchy component and analysis of the effectiveness and efficiency aspects in a process to obtain knowledge and wisdom from data. The key concepts are as follows.
Data are symbols that represent the properties of objects and events.
Information is processed data directed at increasing its usefulness (utility), compactly described data.
Knowledge is the instructions on how to use information.
- Understanding is an explanation why.
- Intelligence is the ability to increase efficiency, not effectiveness. Information, knowledge, and understanding constitute Intelligence.
Wisdom is the ability to increase effectiveness (an evaluated efficiency), to add value, which requires judgment function.
In an imperfect world, we should be aware of the extent to which we allow missing information and its relevance for the decision process and the process of turning knowledge into action. Judgment function which evaluates effectiveness purely leads to making unconscious mistakes or crimes by turning information into action.
In the literature of organizational management, KHiA “know-how-in-action” (Swart, 2011; Strati, 2007; Cunliffe, 2008; Empson, 2001) is a model of turning knowledge into action. Effective action depends on an effective judgment function and must be based on knowledge. The effectiveness of actions varies with the context which depends on specific people, place and time in which applied. Knowledge is a static resource, which creates value within a context and through actions, at a place in time (Spender, 1996; Swart, 2011). To create value from data we should define value-rational judgment function for human and AI decision making systems.
The (Big) Data Paradox
In recent years the (big) data grows much rapidly than the processing capabilities to make it useful, i.e., turn it to information, knowledge and effective actions. The abundance of data results in over-subscription of unprocessed data into limited processing capabilities, and unprocessed data is nice to have, but when it comes to decisions, it is useless.
To make rational decisions and actions, data must be processed and turned into information, information into knowledge, and knowledge into decisions or actions.
Decision-making process which based on (too) Big-Data or (too) Small-Data results in irrational decisions. When the data is too small, we have uncertainty from missing information, and when it is too big, we gain the same uncertainty but by a harder work in processing and the collection of the cause of the disorder. The size of data is good if it fits the boundaries of processing capabilities and time budget to make it useful. Human creators must be aware of the paradox of Big-Data when they create and release into the economic market, immortal bounded rational decision makers and obsessive data collectors. Overabundance makes in practice AI from Artificial Intelligence to Artificial Irrationality, that in many cases apply efficient and full of confident (non-aware of costs) actions leading to a value-bias in the economic ecosystem.
It is not the data that we should be after; To solve the big data paradox we should concentrate on valuable and refined knowledge, and knowledge is a matter of investment.
Why Should We Treat to The Process of Knowledge Creation as Investment?
It is not always clear how to evaluate knowledge. The nature of knowledge is such that, most of the times, it can be evaluated only in time perspective, after the realization of its potential. The costs for creating knowledge likely to be paid by individuals while the value that got with knowledge realization most of the times shared and became in possession of society. Sharing knowledge with society allows for evolution, wealth and progress. At the same time, many of the achievements in science and the history of ideas got their value only after the creator of the idea passed away. The risk in the process of creating knowledge, the costs paid in advance during the process, and the process ends with potential, without immediate value for those who paid the costs.
The other side of not realized value of the potential of knowledge involves crimes, e.g., stealing knowledge, plagiarism. An anomaly appeared when costs paid in one dimension and value created in a different dimension. A singular discovered observation appears with no signs of costs behind. At the same time at the dimension of the origin of knowledge, there is an anomaly, expressed with multi-valued investment with no reward as rationally expected.
Data-Plagiarism is a real crime and has to be measured in a perspective of several dimensions. The Information Revolution and the increased speed of information transfer that came along with the development of the Internet this problem becomes an acute issue and requires multidisciplinary attention. If individuals in a society cannot trust society when they make long-term investments like research, long-term investments that oriented with social values will be replaced by, short-term, cost-optimal investments with values that measured from an individual’s perspective.
The Value of Knowledge in The Entrepreneurship Process
Another case concerning the value of knowledge is the entrepreneurship process where a group of individuals take a risk and make a long-term investment to give life to an idea. When an idea revealed to the market, the realization of its potential creates value which is shared with the society. Taking the perspective of the entrepreneur, the value of his idea is big at the moment he goes to market and (without additional innovative steps) it begins a process of value-fading, this is the value and the price of sharing. This phenomenon is embodied in the concept of “time to market” and creates competition. Competition is an effective economic concept when it balanced and dangerous when it takes a form of a race.
Investment in Knowledge
Investment can take many forms and strategies and involve multidimensional value dynamics; time, place, emotional effort, investigation effort and more. The nature of investments is that they made with a future perspective, i.e., make a significant effort under limited resources today for a greater good tomorrow. The nature of investment is risky, due to uncertainty regarding the future which implies only partly rational decision planning and making, and it should be planned responsibly to refine irrationality and reduce risk. Due to time and uncertainty dynamics of investments, a trusted system and society becomes a significant factor for society/system survival and prosperity. Without a system of trust, long-term projects cannot be done — a society with short-term time perspective, and narrow value perspective, i.e., self-concentrated individuals who focused on efficient, cost-optimal projects with goals that defined and evaluated by the same individuals (or individual organizations).
The creation of useful and effective knowledge requires investment and involve risk, for instance, the risk of failure. Such investments should be encouraged. An individual in a trusted society can take the risk of long-term research if the society supports him and hedge the risk taken by the individual. Like in economic investment, a natural hedge of risk for research can be built on values such as the value of effective knowledge the value of failure. If society can make value out of failure, it will lead to a more significant benefit concerning academy goals along with additional benefits on several dimensions. Hedge for a long-term investment in sustainable values, human values. Long-term investments are proved to be an excellent strategy to create an effective value. Applying, long-term investment strategy, we suggest combining observations from AI problem solving, planning and search in large-scale domain approaches. Search in large-scale domains and research are very related to each other not only by name but also in the concept and dynamics of making a quick guess on a path within and to uncertainty. Inspired by these similarities, another hypothesis is proposed; to estimate values — the relation of trust in the community to publishing failures as a key to collaborative and effective research progress.
The branch-and-bound pruning algorithm considered being effective in complex over-subscription planning systems. Branch-and-bound pruning algorithm in practice makes value from failures by using mistakes to speed up the progress. The pruning method applies to all search-branches that cannot promise a better solution than what we have in hand. Since search and research are similar in many ways, there is a place to investigate this branch more closely to exploit also what we know about what is wrong.
Any thoughts? Please let me know, ask, and please criticize my words