The World Economic Forum is mostly known for its highly anticipated annual meeting in Davos, Switzerland — a scene of global intrigue and a convergence point for the planet’s most influential politicians, investors, activists, CEOs, and economists.

But the work of this prestigious NGO extends far and beyond the confines of a singular event. Established in 1971 as a not-for-profit foundation, the World Economic Forum has made it its year-round mission to “improve the state of the world by engaging business, political, academic, and other leaders of society to shape global, regional, and industry agendas”.

Sustainable AI for a Healthier Future

One of the foundation’s key platforms is “Shaping the Future of Technology Governance: Artificial Intelligence and Machine Learning”. This program brings together key stakeholders from the public and private sectors to co-design and test policy frameworks that accelerate the benefits and mitigate the risks of Artificial Intelligence (AI) and Machine Learning (ML).

As part of the Chatbots RESET project, one of many under the broad umbrella of initiatives launched by this platform, I was fortunate enough to be invited to participate in the Chatbots for Healthcare virtual workshop, a gathering of some of the brightest minds in conversational healthcare, tasked with coming up with a set of answers and ideas to the field’s most pressing issues. On a personal level, this experience was both incredibly enlightening and, in many ways, humbling. But as an extension of Hyro, a conversational AI platform for healthcare, it was as if I came across a treasure trove of actionable insights to be explored and tested for the benefit of our clients.

In the spirit of the free exchange of ideas so brilliantly exemplified by the World Economic Forum, here are some of the key takeaways from the workshop, freshly conceived by conversational AI experts from Google, Microsoft, Babylon Health, and more.

Conversational Health Use Cases

Conversational AI in healthcare (conversational health) offers a wide array of use cases from digital patient access to care management and delivery. Following several rounds of discussion, the workshop’s participants reached a consensus on the five primary applications of conversational health:

1. Diagnostics — In the past year, more than half of the planet’s population at one point during the COVID-19 pandemic, was ordered to stay indoors under strict social distancing measures. Virtual diagnosis and screening tools — up until recently considered merely ‘nice to have’ — have been rapidly and overwhelmingly adopted by health systems across the world.

As many medical centers quickly hit maximum capacity, the urgent need to move the triaging process from the facility to the patient’s home became abundantly clear. Conversational AI, coupled with Natural Language Understanding (NLU) and Machine Learning (ML) capabilities, can conduct a patient’s triage seamlessly and virtually.

Moreover, a virtual AI assistant can aggregate the patient’s information and deliver an instant risk assessment or score, which, with limited availability, automatically places the patient in line for care.

2. Administration — According to a 2019 study published by JAMA, administrative costs are the largest source of wasteful-spending in the American healthcare industry, totaling $226 billion a year. These administrative procedures include appointment scheduling, insurance claims approval, billing, payment processing, and HR management. Health Affairs estimates that for every 10 physicians providing care, almost seven additional people are engaged in billing-related activities. Although much can be argued that the entire system is in dire need of streamlining, conversational AI can, at the very least, fill the gaps and reduce the costs of the manpower currently required to perform these tasks.

3. Customer Service — In recent years, we are witnessing a dramatic shift in the healthcare industry from a B2B to B2C mindset. As almost all verticals, from travel to retail through to education and real estate, are rapidly transitioning to online systems, patients who have grown accustomed to certain standards of service are expecting the same of their healthcare providers.

In each of these fields, conversational AI is already harnessed to increase engagement, conversions, and sales. According to Oracle, 80% of sales and marketing leaders say they use conversational AI in their CX (customer experience) or plan to do so by the end of 2020. To remain competitive and consumer-facing in the “red ocean” that is the healthcare industry, many organizations are now rushing to implement conversational AI as part of their patient engagement efforts. This trend is only expected to grow.

4. Companionship and Care Management — A Pew Research Center study found that In the U.S., 27% of adults ages 60 and older live alone. Approximately 85% of older adults have at least one chronic health condition, and 60% have at least two chronic conditions, according to the Centers for Disease Control and Prevention (CDC).

As the “Baby Boomer” generation — the largest in the history of the U.S. — joins these statistics, dynamic, accessible, and easy to use tools for ongoing care management are becoming a necessity. Conversational AI can make up for the absence of a family member or caregiver, in keeping this vast swath of the population connected, engaged, and safe. Virtual Assistants operating as in-home nurses can keep patients on track of their medication regimen, follow up on recent medical procedures, relay that information back to a human provider, and even act as a companion (albeit limited) for social interactions.

5. Wellness and Nutrition — The global wellness industry is currently worth a staggering $4.2 trillion. Within this gigantic market, healthy eating, nutrition, and weight loss account for $702 billion. As healthcare organizations increasingly diversify their offerings with a strong emphasis on outpatient care, nutrition services have become building blocks for greater profit.

Conversational AI is instrumental in bolstering and supporting custom-made nutrition plans. A virtual assistant specifically tailored to the nutritional needs of a patient can set off alarms and notifications as well as answer questions and, if needed, hand the conversation off to a live representative. Imagine a bi-directional solution that automatically logs caloric intake, or fiber count, and shares that information with a specialist in real-time.

Up-Skill Untrained Health Workers in Call Centers

Won’t this replace jobs? Not quite. Conversational AI, at its best, has the potential to enhance the abilities of its users. One exciting idea raised during the workshop was using conversational AI to fill in any knowledge gaps healthcare call center operators may have.

When considering the fact that there are hundreds of thousands of medical terms, it’s easy to understand why an untrained human operator may find the help of artificial intelligence useful. A virtual assistant will handle the initial part of the conversation, ascertaining the caller’s name, date of birth, condition, insurance, etc. The information collected is displayed in real-time on a designated dashboard for the operator to prepare for the hand-off and jump in at any point if deemed necessary.

What this combination of human and machine adds up to on a larger scale, is the creation of viable employment opportunities for untrained workers in the healthcare sector.

Certified Information Versus the Spread of Misinformation and Fake News

As was mentioned in our recently published COVID-19 Insights Report, In a July 2019 survey conducted by BMC, 72.1% of respondents stated that their general practitioner (GP) was their information source of choice for health-related questions.

Healthcare organizations have a critical, at times, life-saving responsibility to provide their patients with certified, vetted information. This is further highlighted by a tsunami-like proliferation of fake news and misinformation in the wake of COVID-19. In the context of conversational AI, accountability and transparency are foundational to the ethical use and dissemination of information.

As Natural Language and Machine Learning models become more complex and advanced, it is of paramount importance that all medical-related information ingested goes through a meticulous screening and examination process.

Obstacles to a Conversational Future

While most of the Chatbots RESET project focused on the pivotal benefits of the technology, part of the initiative was spent highlighting some existing and potential hurdles involved with the widespread adoption of conversational AI in healthcare.

Several issues, such as miscommunication between chatbots and customers, AI hesitancy or negative customer perception of chatbots, and omission or reduction of customer preferences in interacting with AI versus human beings, took the spotlight. But those weren’t even the most daunting aspects. Concern hit the stage regarding inaccurate/poor guidance, improper diagnosis or screening, and the possible neglect of intervention when necessary.

In order to develop difference-making AI for healthcare, there needs to be emphasis on the trifecta of pillars for widespread governance, listed by the World Economic Forum as “transparency, reliability, and data security.“

According to the World Economic Forum, some of the more classic AI governance gaps include:

1. Validation/Accreditation

• Should chatbots have limitations on the types of use cases they’re deployed in?
• How do we regulate these uses, and are those current measures enough?
• As medical experts must receive qualifications in order to perform their duties, must chatbots also receive some sort of accredited validation before deployment?

2. Performance Guarantee

• How accurate will the systems be in understanding, routing queries, and sourcing validated answers?
• Will systems flag urgent issues or emergencies that are out of scope/ incapable of handling?
• Who governs the oversight of the medical information presented?

3. Patient Expectations

• Will it be made clear to patients that they’re engaging with a non-human entity?
• How will systems handle desired hand-off, if patients want to speak to a human being?
• In terms of access, how simple will it be to connect? What kind of digital literacy is needed?

4. Legality/Privacy/Security

• Who will shoulder responsibility for wrong diagnosis or misdirection or lack of timely response?
• How would consent work for using the system and allowing access to and storage of personal data and chats?
• How will Electronic Health Records play a role in the development of this technology?

Paving the Path Forward

While that list seems lengthy, note that a majority of these gaps are actually already being filled by various players in the conversational AI space. Returning to the previously mentioned pillars of governance, there should be a focus on “transparency, reliability, and data security.“

It’s vital to set expectations for patients, and to divulge the capabilities of the system, at the onset of engagement. For instance, our virtual assistants don’t open with generic text, rather, they announce their explicit purpose, such as the example below regarding finding physicians and helping with COVID-19. Setting expectations early results in less friction between patient and provider, should the patient require other use cases or assistance that isn’t relevant to the designated purpose of the chatbot.

Reliability comes from robust understanding; natural language, a core piece of our technologies, allows for open dialogue for all dialects, limiting bias and creating a wide range of actionable inputs. Confident understanding allows for accurate triggers to exist, so that certain patient intents/phrases will automatically lead to handoff to a human, or at minimum, acknowledgement to the patient that their desired action is out of scope.

With regards to data — heightened encryption, PHI reduction, and the democratization of data between healthcare organizations, EHRs and the patients they serve, are just some of the standards being set.

Beyond the product itself, let’s zoom in on how conversational AI might mistakenly create a “digital divide”, which can exist due to factors including age and socioeconomic status.

Patients who are less tech-savvy, such as the elderly, or those who do not have the financial means to access the mediums that digital care is delivered through, could become neglected. To ensure that advanced digital care such as conversational AI is reaching the maximum number of patients possible, the digital literacy required to use the virtual assistants should be as simplified as can be. To guarantee that the gap is bridged between those who have better access to technology and these types of services, an emphasis needs to be placed on omni-channel solutions.

Conversational interfaces should meet patients on whichever channels they have access to, whether that’s SMS, mobile apps, websites or call centers, so that nobody is left out of the digital revolution.

While there is still progress to be made, the World Economic Forum prioritizing conversational AI as a key point of discussion on the global agenda echoes this technology’s meteoric rise.

We should all be determined to continue to be active members of the conversational AI community, and to contribute as much as we can to chatbot advancement during the COVID-19 era and beyond.

At Hyro, we are offering a free COVID-19 virtual assistant to help health organizations overwhelmed by a spike in traffic. Follow our journey on LinkedIn or Twitter or shoot me an email at aaron@hyro.ai.

The Artificial Intelligence and Machine Learning platform at the World Economic Forum is working on the governance of chatbots in healthcare. For more information on this project or to engage with this project, contact Arunima Sarkar, Project Lead, at arunima.sarkar@weforum.org, or Venkataraman Sundareswaran, Project Fellow, at vsundar@weforum.org

Previously published on: https://www.hyro.ai/post/why-the-world-economic-forum-wants-ai-chatbots-in-healthcare