Humans and AI: AI and Individuality

I started a family late in life. I have a three-year-old son and just a couple of months ago, celebrated my fifth wedding anniversary. There are pluses and minuses to starting late.

My lifestyle choices don’t neatly fit into customer categories that marketers have been using for decades. Some marketing campaigns assume that if I’m a fifty-something man purchasing a product for a child, I must be a grandfather. Other misguided marketing campaigns have tried to sell me shampoo or funerals.

I’m not a grandfather. I don’t have any hair to shampoo. And despite being on the wrong side of fifty, I’m not planning to die any time soon!

I’m an individual. I want to be treated as an individual. And research shows that I am not the only one who feels this way.

Personas

A customer persona is a fictional character created to represent a type of customer. Because not all customers are the same, marketers place them into groups (segments) with similar characteristics and behaviors. Personas are subjectively constructed.

The proposed cognitive benefit of personas is that humans are better at thinking in narratives than abstract data about customers. This approach might help marketers use their imagination and empathy to infer what customers want and need. Although there hasn’t been a lot of research on their effectiveness, the authors of “Real or Imaginary: The effectiveness of using personas in product design” conclude “This study demonstrates the effectiveness of using personas in the product design process and while more research is needed, there is now some objective evidence that using personas does work.”

But others have criticized the use of personas. It isn’t a science. People’s cognitive biases can affect the way they create and use personas. For example, due to confirmation bias, the tendency to interpret and favor information in a way that confirms or supports one’s prior beliefs or values, personas can reinforce outdated stereotypes, hindering the use of imagination and creativity. Because they’re subjective, it’s difficult to know how many users are represented by a persona or whether a persona is relevant for intended customers.

By grouping large numbers of people, personas ignore individuality. We don’t know how much individuals vary from the persona stereotype.

Cluster Analysis

Cluster analysis or clustering is an unsupervised learning method that groups objects in such a way that objects in the same group (a cluster) are more similar to each other than to those in other groups (clusters). Much like personas, cluster analysis is a tool used by marketing teams for customer segmentation and insights into customer characteristics and behaviors. Unlike personas, however, cluster analysis is data-driven. It can quantify the variation between groups and individuals in groups.

In the twenty-first century, due to its simplicity, interpretability, and low computing load, cluster analysis became standard practice for customer segmentation. Because it quantitatively identifies the common attributes of members of a group, it can inspire new persona definitions.

But despite the allure of a data-driven approach, cluster analysis is arbitrary. It can be unstable. Small changes in the data or parameters can result in the creation of significantly different groups. Because cluster analysis is an unsupervised algorithm, it does not attempt to select groups that optimize business goals. Although the groups it creates might have differentiating attributes, those attributes might be unrelated to business outcomes, such as customer purchasing preferences.

By grouping thousands, if not millions, of consumers together, cluster analysis ignores individuality. We don’t know whether the clusters or groups are relevant to critical business decisions.

Individuality and Acceptance of AI

What do we know about individuality and consumer acceptance of AI decisions?

The authors of “Resistance to Medical Artificial Intelligence” explore consumers’ receptivity to medical AI. Although modern medicine is founded on rigorous experimental design and statistical analysis, and many research studies have shown the superiority of objective analysis over human intuition, medical AI adoption will depend on consumer receptivity and trust in this new technology.

The results of previous studies showed that people believe their health-related circumstances and medical conditions are unique and depart from standard clinical criteria more than those of other people. The researchers set out to test whether consumers are more resistant to healthcare delivered by AI providers because of concerns that the AI will neglect the unique characteristics of their case.

In the experiment, college students were offered the opportunity to schedule an appointment to diagnose their stress levels. Each participant received a questionnaire and a test kit to return a saliva sample. One group was told that the data would be analyzed by a physician. The other was told that the data would be analyzed by an AI. Both groups were shown an identical historical accuracy rate for their test provider. The physician and AI were described as having the same accuracy rate. After the participants received this information about the stress testing process, they were asked whether they wanted to schedule an appointment. 

Students were significantly less likely to schedule a diagnostic stress assessment when the provider was automated than when the provider was human. The researchers concluded that “consumers may be more reluctant to utilize medical care delivered by AI providers than comparable human providers, because the prospect of being cared for by AI providers is more likely to evoke a concern that one’s unique characteristics, circumstances, and symptoms will be neglected.”

The researchers noted previous research that shows humans use heuristics that cause them to believe that machines are inflexible and treat every case in the same manner. When an AI system does not signal how it is using a consumer’s unique attributes, “people use perceptual cues and former experiences with inanimate objects to characterize computers as rote, rigid, and inflexible.”

Because a lack of perceptual cues cause consumers to believe AI systems are inflexible and don’t treat them as individuals, it’s important to demonstrate how AI recognizes their unique attributes and treats them as an individual.

In another series of experiments, the researchers gave participants a choice of medical provider but changed the disclosed accuracy levels for each provider. Although it wasn’t clear from the original research paper, substantial analysis of the same data concluded, “People actually did prefer the AI provider so long as it outperformed the human provider.” This gives us another path to AI acceptance. We should use advanced algorithms to ensure that AI systems outperform human processes and communicate this outperformance to consumers.

Individualization

Now that computer power and data storage are more available and affordable and digital transformation is becoming mainstream, best practices are changing. Rather than targeting personas for market segments, it’s possible to treat each customer as an individual, a segment of one, and deliver a unique customer experience tailored just for them.

Hyper-individualization is the new standard for marketers and product designers. But this hyper-individualization cannot be achieved with legacy processes such as personas and cluster analysis. Modern machine learning algorithms use the exact attributes and historical behavior sequences of individual consumers to predict the next best action you can take that will build their engagement with your brand.

Although some businesses are worried about consumer reactions to how their data is used to target them as individuals, if you follow best practices and do the right thing, your customers will appreciate your efforts to treat them as individuals. A recent survey of consumers showed that 87% of customers find it acceptable for brands to use their data to personalize communications as long as it is relevant to them or from companies they have recently purchased from. But beware of selling or buying consumer data. The same survey reported that 67% of customers found it intrusive when brands they had never purchased from used their personal information to tailor their marketing.

Humans and AI Best Practices

Consumers want to be treated as individuals, and the only way to achieve this at scale is through AI. But the research shows that there is a right way and a wrong way to create AI-driven customer experiences. Legacy marketing practices, such as cluster analysis and personas, won’t give consumers the individual experience they’re after. Use advanced machine learning algorithms to power your AI so that your AI outperforms the legacy human processes it replaces. Then clearly communicate that outperformance to your customers. Use prediction explanations to demonstrate how the AI has considered their unique attributes. If you use customer data wisely to value your customers’ uniqueness, they will thank you for it.

About the author

Colin Priest

VP, AI Strategy, DataRobot

Colin Priest is the VP of AI Strategy for DataRobot, where he advises businesses on how to build business cases and successfully manage data science projects. Colin has held a number of CEO and general management roles, where he has championed data science initiatives in financial services, healthcare, security, oil and gas, government and marketing. Colin is a firm believer in data-based decision making and applying automation to improve customer experience. He is passionate about the science of healthcare and does pro-bono work to support cancer research.

Meet Colin Priest