In part 1 of this post (click here if you missed it), we talked about what ‘Artificial Intelligence’ is. We also talked about one of its major uses for nurses – to provide insights and guidance. That is AI’s more obvious use, but what happens when AI takes on the work of nurses more directly?

Here we get into the 2nd major way AI may impact nursing:

AI extending human relationships:

Use 1: AI becomes the nurse’s newest assistant to whom we can delegate tasks.

There are a number of AI applications – some with robots – that can perform assistive work for nurses.  Imagine extending the nursing assistant workforce so no unit is left understaffed. Imagine being able to flexibly add staff who never get tired, sick, or need a break.

Texas Health Presbyterian Hospital in Dallas has had a robot named Moxi on the team for over a year.  

From a company named Diligent Robotics, Moxi is able to help clinical staff with things like fetching supplies, delivering lab samples, and removing soiled laundry.

Instead of a nurse stopping in the middle of a task to get additional supplies, that nurse can simply call Moxi.

Another team member nurses may soon meet is Dragon.  

You may know Dragon from the speech dictation software used by physicians for dictating notes.  Well, like many things, Dragon has upgraded, and now has an AI-powered version called Dragon Ambient Intelligence.  

Dragon Ambient Intelligence can listen in on a clinician-patient conversation and automatically document the encounter.  But more than that, it can act as a virtual assistant creating orders and completing tasks.

Imagine never having to document again…I know a lot of nurses who would rejoice if AI could take over that function.

Use 2: AI helping to reduce with senior social isolation.

Social robots for seniors have already been in use in countries like Japan.  These are robots who can have conversations, display human-like emotions, and provide companionship for the elderly.

While social robots are not yet widespread, COVID-19 and the isolation it caused many seniors may soon change that.

Not sold on the idea that anyone would want a robot in their home?  Well, Siri, Alexa, and Google have already provided AI in the form of virtual assistants.  A robot is the same idea but with a body.

Additionally, the robots are pretty cute.

SoftBank Robotics makes two robots named Pepper and Nao: help with companionship, display human-like emotions.

Nao is designed to be smaller with the intent to be used with children.  For example, there are studies using Nao with therapy for Autism Spectrum Disorder.

Buddy the Emotional Robot is another example. Buddy is designed to be part of the family, and offers everything from home security to games and quizzes.

As nurses, we understand the benefits of companionship for seniors.  The opportunity exists for AI to fill a needed emotional wellness gap. AI can also help the senior stick to treatment plans and remind them to take medications.

AI may become the most reliable historian of a patient’s health when care is needed.

Use 3: AI supporting daily disease management.

A daily call from a health coach would increase compliance with treatment plans, But few organizations have enough staff to provide that level of attention.

AI has the potential to fill the gap between check-ins with a patients health team.  In this space the term Virtual Nurse describes a virtual image of a nurse who speaks with the patient daily.

The company Sensely creates virtual avatars that can talk patients through tasks like gathering vital signs and other biometrics.  The virtual nurse can pull data from bluetooth-connected devices to add to the patient records, and is capable of conversation.

Think of it as a chatbot but with a face and human-like expressions.

And, like other examples of chatbots, the virtual nurse can determine when to connect the patient with a human clinician.

Tavie is another company that produces a virtual nurse. But this one uses video clips of a real nurse.  Tavie is meant to improve treatment adherence over a period of time. The makers have a list of disease applications, from diabetes to transplants.

Here is an example of the Tavie virtual nurse at work supporting patients with HIV:

Challenges with AI:

All technologies have their challenges, and artificial intelligence is no exception.  Let’s take a look at each challenge, real or imagined.

Challenge 1: Perception AI is here to take all jobs away from humans.

Let’s be clear: for nurses this challenge is really more imaginary than reality.  While it is true for certain types of jobs, nursing requires many skills machines cannot easily copy.

According to a recent study by MIT economist Daron Acemoglu, the hardest hit will be manufacturing jobs.  Areas like automobile manufacturing, electronics, and plastics production most easily incorporate robots.

Another analysis of jobs done by a team led by IBM’s chief economist, Martin Fleming, found the impact of introducing AI has made ‘soft skills’ more desirable in the job market.  

The ability to think critically, use judgement and reason, and employ creativity when needed – AI is not able to copy these traits.

Challenge 2: AI carries bias.

Unfortunately this challenge with AI is all too real. The racial and gender biases that exist in humans translate into machines.

Recent headlines of IBM, Microsoft, and Amazon halting the use of their facial recognition technology by police departments brought racial bias in the software to national attention.

But how can a machine be racist or sexist?  

Keep in mind that these machines are programmed and trained by humans using data created by humans.  Along the path of creating AI there are several points where racism and sexism can enter. One place is through the selection of data.

MIT researcher Joy Buolamwini provides an example of an experiment she did. She wanted to see how well facial recognition software could identify gender:

This software was so inaccurate for dark skinned women, it was no better than flipping a coin.

Would you want AI to say you are the suspect in a crime if it is only as accurate as a coin toss?

This could be a big problem in healthcare. Psychology and Social Work commonly use clinical notes.  Machine learning and natural language processing can interpret the information from these notes.  

However, that also means racism and sexism baked into those notes carry over to what machines picks up.

AI trained on notes from a social worker who regularly describes female patients as ‘dramatic’ and ‘hysterical’ would miss anxiety disorders in women.

There is still much work to be done in this space, and nurses need to be aware of this issue.  Assuming that a machine cannot be biased misses the fact that it was created by humans.

Challenge 3:  AI is only as good as the data it has.

There is still much excitement about the promise of ‘Big Data.’ But much of healthcare data sits in silos.  Regulation limits what can be shared freely. There are also some populations who do not have access to the tools that would capture their data.

For example, if data from Fitbits becomes a normal part of heart treatment, anyone who does not have one will be at a disadvantage. AppleWatch wearers would have less personalized cardiac treatment options.

This ties to the challenge for nurses to be aware of the limitations of data fed to AI.  We should question where the information came from. We should also provide feedback on how to get the most accurate and representative sets of data.

Tips for Nurses working with AI

Rather than taking away the job of the nurse, I believe AI will transform it.  

Nurses will continue to be the integrators of information. But we will have more intuitive tools that draw from a larger amount of data.

How can nurses best prepare for working with AI?

1. Keep skills up to date: AI may enable nurses to be more autonomous and perform diagnostics and treatments currently requiring advanced degrees.  To be ready for this responsibility, nurses will need to invest in continuing education and constantly refresh our skills.

2. Practice critical thinking: AI should be 1 of many tools the nurse has available. But it should not replace our ability to think.  We also know that AI can have bias built in, so we need to continue asking questions.  We do not want nursing to fall into the same traps of the justice system where judges allowed biased algorithms to overrule their experience and intuition.

3. Listen to patients: AI will allow patients to be more informed and in control of their care.  The days of telling a patient ‘ignore what you read online’ will fade.  AI will not only have more information about that patient before they ever see a clinician. It will also suggest a probably diagnosis.

4. Insist on a seat at the table: AI is still created and deployed by humans.  Nurses must insist on a seat at the table to be part of decisions on AI use.

Even when AI carries the title of ‘virtual nurse,’ it does not actually replace nurses.  It helps nurses to do more for our patients, understand our patients better, and extend our connection with them.

We must insist on being part of the conversation and not a mere bystander to it.  Without vision and guidance from nurses, AI can add to problems instead of solving them.