Machine Learning Algorithms in Business Applications

Machine Learning Algorithms in Business Applications:

What Every CEO and GC Needs to Know

By Allison Hushek on July 18, 2023

Machine Learning (ML) algorithms can be a plaintiff’s lawyer’s dream. As CEO and GC, there’s a lot you need to know about implementation of ML algorithms in order to advise the C-Suite, the engineering team and employees who utilize the results.

Business Applications. Before we talk about risk management, it’s important to understand business applications of ML. Algorithms can be deployed across various departments to benefit a company. Is your business trying to improve the customer journey by making sophisticated predictions for similar product purchases? Is it trying to reduce fraudulent transactions? Is the company attempting to elevate the employee experience or smooth the hiring process? Is it trying to make its targeted advertising more efficient?

Getting Started. Recent advancements in ML have disrupted industries in such positive ways that every company is on a mission to incorporate algorithms. To help your company get started, an ML Council should be formed, consisting of various C-Suite executives and employees who will implement and utilize the results. The Council should establish the corporation’s goals, and the data collection necessary to achieve those goals. Start with quick wins on short-term projects where gains can be seen in 3-6 months. This will allow in-house teams to build skills in data gathering and results analysis. Then add long-term projects, such as rethinking a company’s end-to-end processes. Identify the company’s values before the project begins, and check back throughout the development phase to confirm those values are in line with the project outcome. Know that limited data and tacit knowledge are not a good recipe for ML. Volumes of historic data are necessary for training and testing. About 30-50% of data used to build the algorithms must be reserved for testing the algorithms. Leave the types of algorithms used to your skilled engineering team (e.g., decision trees, random forests, neural networks, deep learning).

Productivity Increases. ML tends to shine in process innovation, and studies show 7% productivity gains for companies. Product innovation efforts using ML, however, have not been shown to increase productivity. Studies also indicate a 3% increase in patent productivity for a company, which can translate to 4 new patents per year depending on the size of the company. ML is great at finding re-combinations of patterns versus finding new patterns altogether - and at an exponentially faster rate than humans. Robot adaptation by companies has shown 10 times productivity – that means 10 times the value of a company. Despite recent media articles, robots are not out to steal jobs. Research shows robot implementations tend to increase employment, with a nominal decrease in mid-skill management workers with trade degrees or 2-year certifications.

Risk Management: Privacy. Now that ML goals are established at your company, it’s critical for CEOs and GCs to know how to comply with various laws and regulations in order to minimize lawsuits. Privacy laws come into play at the early stages of data collection. Compliance with the General Data Protection Regulation (GDPR) in Europe and the California Privacy Rights Act (CPRA) is key to a successful start. If you’re not a Certified Information Privacy Professional (CIPP), at a minimum you should be aware that privacy laws center around the protection of one’s personally identifiable information (PII). In sum, a person has a right to request what PII is being collected about him/her by the company, as well as the right to request erasure of such data (with some exceptions). Lastly, be aware that the Children’s Online Privacy Protection Rule (COPPA) applies to PII collected on individuals under 13 years of age.

Risk Management: Biases. Next, CEOs and GCs must keep an eye out for biases created by ML algorithms which can lead to discrimination against specific groups or individuals. This occurs when an algorithm delivers systematically biased results as a consequence of erroneous assumptions in the machine learning process. In other words, if the machine is using biased historical data, it will produce biased results. This often shows up in human resources. For example, a machine may recommend white males be internally promoted at a company over other equally qualified individuals because the historic data indicates white males were promoted more often in the past due to biases. Along those lines, a machine may suggest white males be hired for jobs over those equally qualified in underrepresented groups because historic data contains such biases. As another example, a machine may deliver ads for higher-paying executive jobs more often to white males due to biases in past data (see Carnegie Mellon University study in 2015 where Google delivered high-paying executive job ads to the male group 1,852 times in comparison to only 318 times to the female group). Know that a machine learning system may potentially pick up on statistical connections that are considered ethically inappropriate or illegal and replicate this pattern. Identifying ethical guidelines is important in ML implementation, and as noted above, it is recommended that a company identify its corporate values before the project begins.

Conclusion. A company must: (i) allow users control over their data, (ii) have transparency when it comes to algorithms, and (iii) implement auditing procedures to ensure legal and ethical compliance. New York City recently passed an ordinance relating to the use of AI-powered hiring applications (Local Law 144 of 2021, effective July 5, 2023). The law specifies requirements for audits before and after application deployment, and for notifications to the public throughout the implementation process. Penalties for violations in NYC start at $500 per violation and increase to $1,500 per violation. It is anticipated that other government bodies will pass similar laws. A special thank you to the Wharton School for the information I obtained as a basis for this article while earning my specialization in AI For Business through Coursera.

Playbook Law is a solo firm providing General Counsel Services for Startups and Interim In-House Counsel for Established Companies in the Technology, Gaming, Entertainment + Sports industries.

This blog is provided for information purposes only and does not constitute legal advice and is not intended to form an attorney-client relationship.

©2023 Playbook Law, PC. All rights reserved.