Human errors in data analytics, and how to fix them using AI

The benefits of analytics are well-documented. Analytics has helped organizations transform retail experiences, map pathways for trains and trucks, discover extraterrestrial life, and even predict diseases. However, over the past few years, organizations across the globe have wrestled with just how much human error has permeated their analytics attempts, often ending with disastrous results. From crashing spacecraft to sinking ships, transferring billions of dollars to unintended recipients, and causing deaths due to overdose of medication, human error in data analysis has far-reaching ramifications for organizations.

The reason for human error in data analysis could be many, such as lack of experience, fatigue or loss of attention, lack of knowledge, or the all-too-common biases in interpreting data. However, what's common among these errors is that they are related to humans reading, processing, analyzing, and interpreting data. Artificial intelligence can effectively combat human error by taking up the heavy lifting involved in parsing, analyzing, drilling down, and dissecting impossibly large volumes of data. It can also perform high-level arithmetic, logical, and statistical functions at a scale that would otherwise be impossible by human-led, self-service analytics alone.

Below are five of the most common human errors that can be eliminated using AI:

1. Confirmation bias

It's easy to spot a yellow car when you're always thinking about a yellow car. Confirmation bias impacts the way we search for, interpret, and recall information. In the business world, gut instinct quite often trumps data; and data is manipulated, omitted, misrepresented, or misinterpreted to concur with one's own beliefs. And in cases where data doesn't concur with beliefs, the information is faulted and disregarded. Artificial intelligence eliminates this way of cherry-picking data because it uses historical data to look for trends, patterns, and outliers, thereby providing accurate, bias-free results.

Lockheed Martin, one of the world's foremost aerospace companies, uses historical project data, also called dark data, to manage its projects proactively. By correlating and analyzing hundreds of metrics, the company was able to identify leading and lagging indicators of program progress, predict program downgrade, and increase project foresight by 3%.

2. Inability to break silos

Far too many organizations struggle with data-related issues such as organizing multiple sources of data, a lack of collaboration between data sources, low data accuracy, and poor data accessibility. Artificial intelligence can easily break silos by communicating with and correlating large data sets from several applications, databases, or data sources using relational data modeling techniques.

Recently, multiple state governments in India decided to collaborate with the National Green Tribunal on Project Elephant—to assess and prevent elephant deaths on railway lines connecting multiple states—after The Hindu, a leading national newspaper, published a report highlighting the time, frequency, and common routes in which elephant deaths frequently occur. The newspaper was able to put together this report by collating data from railways and the forest reserve departments.

3. Downplaying losses

It's human nature to be loss-averse. Toyota downplayed the impact of faulty brakes in its cars, resulting in some Toyota models being pulled off Consumer Reports’ list of recommended vehicles. BP downplayed the impact of the Gulf of Mexico oil spill by putting out polished ads apologizing for a “minor spill,” until it received severe backlash from then-President Barack Obama, who said the company should have used its PR budget to clean up the spill instead.

Downplaying loss creates tunnel vision and incapacitates leaders from making effective decisions. And in the long run, this can prove costly for the organization. Because of artificial intelligence's analytical DNA, it understands and interprets data as it is and doesn't favor positive trends over negative trends, unarguably eliminating the human tendency to favor positive outcomes. This makes AI-driven analytics an ideal ally for leaders looking to make decisions based on complete facts rather than a partial picture.

4. Inflated predictions

Another downside of human-led analytics is the habit of presenting inflated predictions of the future. Be it forecasting budget requirements for the organization, predicting property damage after a natural disaster, or predicting a fiscal deficit or inflation rates, humans tend to inflate predictions based on their own assumptions and experiences. On the contrary, AI-led analytics tends to be more accurate because it makes predictions based on driving or arresting forces and external or environmental stimuli. The US Navy leverages artificial intelligence and machine learning to predict part failures proactively and plan preventive maintenance of its aircraft and ships. This enables sailors to spend more time focused on missions and less time on repairing aircraft when they fail.

5. Inability to go beyond surface-level analytics

Drilling down to analyze the root cause of problems can put businesses light-years ahead of others that do not follow such practices. Root cause analysis can identify agents causing a problem, hint at remedial measures, and offer ideas to prevent such problems in the future. But with too many data sources, structures, and silos, it becomes impossible for humans to collate, analyze, and drill down to perform root cause analysis. AI-driven analytics can bypass these barriers by effortlessly digging into multiple levels of data simultaneously. Additionally, AI can also overlay several possible scenarios to come up with the most probable cause of a problem.

It's the age of AI

The benefits of AI-driven analytics are many, from providing actionable insights in minutes to eliminating errors or biases in self-service analytics. Now that more and more business leaders are turning to AI to get insights that propel their business, we can expect to see growing adoption of AI in analytics in the Middle East and globally.

Note: This blog was originally published in Gulf Business.