Predictive Analytics

Predictive Analytics: Inspiring and Terrifying Ethical Questions Posed by AI

By Ardeshir Ghanbarzadeh
Share on LinkedIn Tweet about this on Twitter Share on Facebook

The advent of cutting-edge technologies rooted in artificial intelligence and machine learning are changing the way we work and live. Narrow AI technologies such as Amazon’s Alexa have begun to alter how we interact, communicate, shop, travel and entertain and will continue to advance to more sophisticated use cases with time.

Analysts say the fourth industrial revolution, Industry 4.0, is already underway as modern smart technology begins to automate traditional manufacturing practices. Industry 4.0 is automating jobs and moving workforces around, forcing humans to adapt as computer power can now assist with cognitive tasks at work.

However, it cannot be overstated that even the best technologies in the world won’t have value without careful considerations of both its benefits and disadvantages.

What are Predictive Analytics?

Predictive analytics uses a combination of data sets from multiple sources, statistical analysis of the data to find relationships and correlations, develop predictive algorithms and models, machine learning and artificial intelligence to more accurately predict future probabilities of a given outcome, uncover trends and discover new opportunities.

Companies collect lots of data, but tasking a human to sift through data in search of actionable information often isn’t practical. Instead, software can be used to provide humans with answers to what would likely happen given current data, to help guide which actions they should take that could affect that outcome.

<<To learn more, the Logi Analytics BI Encyclopedia offers an introduction to predictive analytics.>>

Predictive analytics uses a combination of data sets from multiple sources, statistical analysis of the data to find relationships and correlations, modeling, machine learning and artificial intelligence to more accurately predict future probabilities of a given outcome, and uncover trends and opportunities.

Data analysis has expanded from select highly trained roles to every level of an organization, including non-technical users. This means organizations now have the flexibility to use data in workflows for a range of functions to increase operational efficiency, user productivity and confidence in business outcomes.  Industries deploy predictive analytics to support use cases such as fraud detection in insurance, reducing risk in financial services, determine vaccine efficacy in healthcare, managing inventory in supply chain, understanding consumer behaviors in retail or fortify computer networks in cybersecurity. Within B2B or B2C sales, organizations can use predictive analytics for revenue forecasting, “what if” scenarios, and cross-sell and upsell opportunities.

Using Predictive Analytics in Automation

While automation has and will continue to be used in all industries increase production and scale, applying predictive analytics to workflow automation is more complex. Even though software continues to push the boundaries of what is possible, predictive analytics is about steering decisions made by people towards a desirable outcome, and not about executing a narrowly defined, repetitive task.

The level of certainty of a machine derived prediction is a function of the completeness and accuracy of the data used to design and test the predictive models. For example, census data could be used to identify patterns and develop a predictive model for consumer vehicle purchase decisions based on income, geography and other demographic factors, it will not be able to predict the make, model or consumer color preferences if the census data is not combined with historical vehicle sales data when designing the predictive model.

Ultimately, predictive models are going to give you exactly what you asked for based on how you trained them, so it comes down to vital questions such as where and when do we want a computer to make a decision that can have a large positive impact. There are no easy answers, and this should be an open discussion among business and technology leaders and experts to consider the implications, challenge assumptions and determine how this will change the way they serve their customers.

Technology’s Societal Ramifications

Humans are better suited for some tasks, while AI and software can be used to augment other skillsets. Augmented analytics use machine learning and AI to aid with data insight and analysis to improve workers’ ability to analyze data. And that’s generally okay, but there are still outstanding moral and even potentially legal issues that have yet to be addressed as we look more and more at software to help us make decisions.

If humans allow technology to automate decisions, there should still be human involvement to maintain boundaries on factors that are not easily quantified through data, along with moral or ethical biases which aren’t easily programmable for machine understanding such emotion, mood, or relationships.

Staying with our vehicle example, imagine a scenario where a software programmer needs to instruct a computer to decide if an autonomously driven car heading into an impending accident should save the lives of its passengers or pedestrians outside the vehicle. Who gets saved? How does this get prioritized? Is it based purely on number of survivors? Age? Does the vehicle owner get a say? These are important questions that developments in technology force us to consider. And that’s just one example.

It will take time for legislators to catch up on creating regulatory laws, so it is incumbent on the industry leaders and decision makers to start the debate and do so in an open forum where there is expert representation from the public at large. However, with a so-called digital arms race to stock up on as much AI and deep learning as possible, now is the time to have the hard conversations.

Final Thoughts

Embedded predictive analytics tools are an important component in the ongoing changes to traditional business applications. New technologies are all about enabling human beings, and better business-driven decisions. At the same time there are broader considerations that need to be factored into ensuring these advances have both a positive impact on business outcomes as well as societal outcomes.

There is no question that a lot of thought should be shared in this ongoing industrial revolution, with predictive analytics able to make use of machine learning and AI to power decisions.

Originally published December 14, 2021

About the Author

Ardeshir Ghanbarzadeh is the Director of Product Marketing at Logi Analytics. He has over 20 years of experience in product marketing, product management and customer success and has held positions at Epitiro, Spirent Communications, Metrico Wireless and VoIP Logic where he created products and services for companies such as Panasonic, AT&T and QUALCOMM.