As the field of business analytics rapidly expands to include new technologies, techniques, and deliverables (e.g., predictive and prescriptive modeling techniques, Big Graph and Big Data Analytics, etc.), stakeholders throughout the organizational strata face a significant challenge – recognizing and mitigating the effects of bias in a growing field of data and information. The means of addressing this challenge include but are not limited to the use of known statistical bias management techniques, reliance on facts, recognizing personal preferences, and avoiding overreach.
Known Statistical Bias Management Techniques
Bias is a “a tendency to favor one side too much” (Thorndike Barnhart, 1997). “Tunnel thinking” is an expression of bias wherein “once we arrive at a conclusion, all subsequent information tends to reinforce it” (Bartlett, 2013, p. 62). Bias can be managed in statistics through several approaches, including the calculation of standard deviation wherein the divisor is reduced by one to increase the value of the resulting quotient, thereby artificially inflating the value to accommodate the bias inherent in a given sample. Bias is more difficult to mitigate with a management team, for example, when expertise is lacking. “When bias arises from what the informant knows and does not know, the problem is not motivation but limited expertise. Further, these limitations are not salient to us, because we simply ‘do not know what we do not know’” (Drinkwater et al., 2018).
Reliance on Facts
“In the absence of facts, we are forced to depend on opinions, which make us vulnerable to overconfidence and other sources of bias” (Bartlett, 2013, p. 50). While an organization may be rife with data, without the appropriate data analytics leaders, analysts, and technology to cultivate and harvest this resource, the data cannot be mined or applied in an effective manner. In this instance, opinion, not fact, seems to rule the day. At times, the prevailing or consensus opinion may become codified and accepted in an organization or a society, irrespective of if it is only being endorsed by a small subset of the population and even if it is incorrect and flies in the face of known, documented factual accounts.
This is not a new phenomenon and is reminiscent of some of Jesus Christ’s encounters with the Pharisees, Sadducees, and the Sanhedrin. Jesus always points to the truth, and in these cases, He often cited the Holy Scriptures because “all Scripture is breathed out by God and profitable for teaching, for reproof, for correction, and for training in righteousness” (English Standard Version Bible, 2001, 2 Timothy 3:16). This can be observed throughout the New Testament, for example, at the Sermon on the Mount where He referenced Exodus 20:13-14, the cross where He quoted Psalm 22:1, and even as a boy in the temple where “all who heard him were amazed at his understanding and his answers” (English Standard Version Bible, 2001, Luke 2:47). The Scriptures provided the facts, the recognized standard of truth, and Jesus readily applied this truth in His arguments against the “synagogue of Satan” to further spread the Word of God (English Standard Version Bible, 2001, Revelation 2:9).
In contemporary organizations, one primary selling point for fact-based decision making is that it “is faster”, but to achieve this speed, “we need to simplify the reliable information” (Bartlett, 2013, pp. 50, 59). This includes the content of the data, it’s structure(s), the mathematical techniques applied, and the way the resulting information is interpreted and presented. Big Graph Analytics provides “techniques that are very useful in understanding the structure of a large network and how it changes in different conditions, identifying closely interacting subgroups, and finding paths between the pairs of entities that satisfy different constraints” (Ahmed & Pathan, 2019, p. 179).
Recognizing Personal Preferences
Part of the solution to reduce the effects of bias includes training and education for leaders and analysts (collectively referred to as analytics professionals) alike. “Analytics professionals are trained to accept their ignorance, value humility in presenting results, and qualify their statements” (Bartlett, 2013, p. 68). Performing an internal customer or stakeholder analysis to better understand the analytics customers in the organization is another approach, and this allows analytics advocates to integrate the preferences of these individuals, especially decision makers, into practice when possible.
Beyond the address of personal preferences, according to Bartlett (2013), decision makers are not utilizing the information available to them because it is not in a format that’s simple and easy to use. “We want decision makers capable of sorting through the chaotic information and disinformation overload, an exercise that can be so overwhelming” (Bartlett, 2013, p. 51).
“We want to be careful about inferring causality…an overemphasis on causal explanations to the point that even random chance has to have an explanation” (Bartlett, 2013, p. 60). Experienced analysts and leaders in a given industry must exercise caution, because knowing too much can be a source of bias as well. “An overemphasis on industry knowledge crowds out good analytics” (Bartlett, 2013, p. 62). The relationships between variables may be appreciated from an experiential perspective but understanding the nature and strengths of the variables and their relationships in any given model or scenario requires data analytics, typically in some form of regression.
“Data analysis provides new insight into the business” (Bartlett, 2013, p. 62). One obstacle to the effective and efficient use of data analytics in supporting decision making is bias. However, the practice of analytics itself can help to reduce the impact of this negative influence. “Analytics facilitates what econometricians call ‘creative destruction.’ This involves challenging the current industry knowledge, validating the more accurate legends, and replacing the fairy tales” (Bartlett, 2013, p. 74). Complementing the practice of analytics with the use of known statistical bias management techniques, a strict reliance on facts, recognizing and integrating personal preferences, and avoiding assumptions and overreach helps to mitigate the effects of bias and bolster the positive impact analytics support, especially in the form of improved decisions.
Ahmed, M., & Pathan, A. (2019). Data analytics. CRC Press. https://doi.org/10.1201/9780429446177
Bartlett, R. (2013). A practitioner's guide to business analytics: using data analysis tools to improve your organization’s decision making and strategy. McGraw-Hill.
Drinkwater, K., Dagnall, N., Denovan, A., Parker, A., & Clough, P. (2018). Predictors and associates of problem–reaction–solution: statistical bias, emotion-based reasoning, and belief in the paranormal. Sage Open, 8(1). https://doi.org/10.1177/2158244018762999
English Standard Version Bible. (2001). ESV Online. https://esv.literalword.com/
Thorndike, E.L., Barnhart, C. L. (1997). Bias. Thorndike-Barnhart Student Dictionary (updated edition).