AI Special Feature: In Pursuit of Artificial Empathy
By Steve Tulk
Most of us understand the importance of empathy. When we acknowledge how patients feel about their health conditions, they’re more likely to:
– Trust their providers,
– Adhere to their treatment plans,
– Improve their health outcomes, and
– Develop loyalty to specific brands.
Despite the benefits empathetic responses can deliver, our industry has yet to fully embrace its integration in the digital space. Other industries, on the other hand, have already forged ahead, executing AI initiatives that had begun years earlier.
Ford, for example, anticipates that many of the vehicles they’ll offer in 2022 will have some form of artificial empathy embedded into their products1. By using biometric sensors and cameras, emotion-detecting cars can tell if a driver is too stressed, too tired, or too ill to drive. Mars Inc. is applying empathy in some forms of product promotions2. By testing facial reactions and emotional responses to promotions (for items such as chocolate, gum, or pet care), marketers are hoping the results can be used to predict short-term sales.
Pharma, on the other hand, has yet to incorporate artificial empathy in physician or patient-focused initiatives. Why is pharma lagging behind? Part of the problem lies with the volumes of data that the industry has collected. While 75% of pharma CEOs agree data is important, according to a PWC report, many say they don’t know how to turn all the data they’ve gathered into information that’s insightful and impactful3. Ouch.
In addition to the data problem, some CEOs believe the industry is simply not placing a higher priority on their AI transformation. A survey of 50 healthcare CEOs from Emerj Artificial Intelligence Research shows that two primary reasons companies are slow to adopt AI are because4:
– They need to be convinced further of the ROI
– Their companies currently lack the skills and resources to turn AI products into reality
Enter machine learning, a subset of AI that is revolutionizing the industry. We can now collect, tabulate, and analyze rich data at a speed and scale that human minds cannot. The ability to handle massive volumes of data quickly and easily has led the industry to explore a wide range of applications for data. Pharmaceutical companies like Pfizer are applying IBM Watson’s machine learning technology to improve its drug discovery process. The hope is that by combining information from various data points (e.g., literature, genomics, biological assays) Pfizer can accelerate its drug discovery and development pipeline5.
Pharma giant Hoffmann-La Roche AG has bought health tech company Flatiron Health to accelerate its cancer research and improve patient care5. Client Novo Nordisk is exploring how AI can improve its digital ecosystem by analyzing multiple sites and thousands of unique pages that yield hundreds of thousands of insight combinations.
Granted, pharma’s move to AI adoption was years in the making, but any digital advances that companies take at this juncture is good news6.
Insights are not Empathy
Let’s be clear. Uncovering insights behind data is not empathy. Artificial empathy (AE) is the crucial next step, and it could lead to AI’s next evolution. Just as there are many ways to feel, there are multiple ways to experience empathy, a complex emotion that takes three forms:
– Cognitive empathy, the ability to understand how someone feels by making some educated guesses.
– Emotional or affective empathy, which requires learning how to listen then relate (“I feel your pain”).
– Compassionate empathy, the act of doing; that is, “what can I do to help?”
An AI with the capacity for empathy could provide more natural interactions, while making judgments that take mood or feelings into consideration. In pharma, that’s like giving brands superpowers to get patients to trust you.
One of the reasons the empathy emotion is so difficult to duplicate is that everyone experiences it differently. What a fibromyalgia patient understands as an empathetic response (e.g., understanding the need to exercise) may be seen as an insult to someone with Type 2 diabetes struggling with weight loss. Bring up the topic of drug pricing, however, and both patients may experience similar concerns, but responses to their anxieties will differ significantly.
Building Empathy-Enhanced Systems
To achieve artificial empathy, machine-learning systems would need to be able to recognize data patterns that are associated with the three forms of empathy. Emotional recognition is an easier problem to solve because machine learning can be used to recognize patterns that are associated with emotion. The information can be gleaned from word usage, facial expressions and body language seen on videos, and voice inflection in audio7.
Luckily, online health-focused communities, social channels, and blogs are filled with these types of data patterns. Platforms like “Care Opinions” give providers, patients, and caregivers a place to share their stories of care. Research-driven communities such as “Patients Like Me” collect and share patient feedback about the therapies, side effects, and disease progression. Disease-specific blogs like “Diabetes Daily” not only provide educational resources, but also encourage “residents” to share their challenges and experiences with others. Social channels such as Facebook, YouTube, and Instagram enable patients to show their visual and audible expressions.
These health-focused communities are fodder to AI’s empathy-enhanced systems. The data patterns could be pulled from these communities and programmed as part of an appropriate response to a specific question. The result is a unique, personalized interaction that could only be seen as empathetic.
It’s important to point out that effective empathy-enhanced systems are only as good as the information that it’s been given. To reduce the risk of developing an empathetic response with a skewed point of view, a diverse group of data scientists will be needed to teach AI artificial empathy8. Diverse teams with different viewpoints and experiences lead to better results.
Why AE is Good for Pharma
It’s been shown in multiple studies that empathy is good for business7. When Johnson & Johnson stepped in to ensure the safety of consumers during the Tylenol tampering crisis in the early 1980s, sales rebounded within a year and Tylenol became the nation’s favorite over-the-counter pain reliever.
In pharma, artificial empathy has the potential to help companies do good:
– Applied in market research, artificial empathy can capture patient sentiments that are not consciously expressed.
– Analysis of real-time expressions from brand campaigns could help marketers better understand patients’ emotional engagement with their digital content.
– Artificial empathy also could be integrated in palliative care to help patients deal with their pain and/or anxieties.
Until artificial empathy becomes a reality, our industry will continue extrapolating the millions of data patterns from digital content to develop information that patients are seeking. But soon, even that many no longer be enough. Patients and healthcare providers want more. They want brands to understand their plight and respond empathetically.