Explainable AI Part II, or: How I Learned to Stop Worrying and Trust the Machine

By: Andrei Bizovi, Senior Business Analyst - Plexure
Published on June 21, 2021

In Part I of this series, we covered what Explainable Artificial Intelligence (XAI) is and why it’s an important consideration for retail decision makers and consumers alike. AI is so pervasive in our daily lives that it’s easy to forget the extent of its influence – from the videos we watch to the music we listen to and the products we buy – and anything in-between. In fact, over three quarters of e-commerce businesses used AI in some capacity for their marketing and sales processes from 2018. Consumer sentiment reflects this: In a 2020 survey by the European Consumer Organisation, over half of the respondents stated that they believed companies use AI to manipulate consumer decisions.

As AI becomes more entrenched and accessible, it also becomes more prominent in public, private and regulatory consciousness. The challenge, then, is to serve models that are not only effective but also transparent enough for both consumers and companies to “get” them, all the while adhering to tomorrow’s (today’s is already too late) expectations around privacy and ethics.

The challenge, then, is to serve models that are not only effective but also transparent enough for both consumers and companies to “get” them, all the while adhering to tomorrow’s (today’s is already too late) expectations around privacy and ethics.

Which brings us to the topic at hand for the continuation in this series: How can XAI be used to improve trust between consumers and companies, and companies and AI service providers? Part I introduced the concept of Black Box models, which are often unapproachable and undesirable for companies to adopt due to their inscrutable nature. The field of Explainable AI is still nascent and actively researched, and while many great minds are focused on the academic aspects of XAI, this article will touch on the business and market implications by looking at the latest literature and relevant industry examples. It will also cover how Plexure addresses the issues of XAI and trust.

The importance of explainability

The demand for XAI is greatest in high-risk industries, where an AI model’s outputs have the potential to significantly impact the individuals involved. The healthcare, insurance and autonomous vehicle industries are clear and obvious stand-outs. In these industries, the process through which outcomes are derived mustn’t only be interpretable and transparent enough to ensure that said process is bias and error free; it must also be able to justify its outcomes from both a legal and ethical standpoint. For example, a hypothetical system that can adjust health insurance premiums based on a customer’s known data must be accountable to customers that were impacted by its decisions as well as assessors and legislators.

While the industry in which Plexure operates, mobile marketing and customer loyalty, is largely free from such high risk situations, explainability still plays an important role. AI-derived recommendations on offers and products that are relevant and individualised to each user are fundamental to our product offering.

When it comes to recommendations, users generally prefer transparency. Not only that, but users should be in command of their own data and any transactions (where consent is given) utilizing their data should be conversational. That is, information should flow both ways and be easy to understand.

Tintarev et. al. lay out some of the design guidelines for well-designed recommender systems, listing out the following criteria:

Imagine a recommender system which fulfills all the above criteria: Turnkey and ready to turbo-charge the marketing efforts of those in high frequency, repeat purchase retail environments like QSR, fuel and convenience, and grocery. A system with explainable logic which can show a marketer how it determined an appealing offer for any given customer at a particular point in time. And a system that drives incremental sales by reducing the reliance on mass offers by pinpointing the most timely and relevant offers to drive the desired outcome. Whether that’s enticing a one-off shopper to return; increasing traffic during a slower time of day; or converting a larger percentage of over the counter orders to drive-through.

Plexure’s mobile engagement engine satisfies all of the above criteria, taking the complexity and guess work out of AI and ML, with user friendly dashboards pulling all necessary data through in easy to understand and sharable reports in various formats. It’s the perfect combination of complex automation coupled with a simplified, user friendly interface.

Plexure's mobile engagement solutions combine consumer, contextual, behavioral and sales data with AI/ML to create next generation marketing campaigns that drive traffic and increase sales. We help some of the world's best-known brands unlock the power of their customer data to eliminate friction and drive desirable outcomes along the entire customer journey, from activation to growth and win-back. If you would like to find out how we could tailor a solution for you, book a demo.

Sign up for our
monthly newsletter

You may also be interested In…

man with cellphone
Explainable AI, Part 1: What it is, why it’s important and the impact it has on business and society

This three-part series delves into their learnings, insights and tips on the subject of Explainable AI.

Increasing customer engagement — the importance of timeliness

As brands fight for market share, consumers are bombarded with marketing messages. So how can marketers cut through the noise and build mutually beneficial relationships with their customers? The key is timeliness.

Read More
Woman eating burger
Loyalty with a capital “L” (and where loyalty with a little “l” fits in)

A brand’s ability to attract and retain loyal customers has never been more important. In this blog, we discuss how loyalty programs can help build a more loyal and profitable customer base.

Read More