Every day during the Covid-19 pandemic US governors emphasize the need to use data to combat the spread of the virus. They look closely at various simulations to predict such things as the peak of the outbreak by location, the potential rate of recovery, death rates, and which works better, lockdowns or extensive social distancing. Such predictions and the ability to analyze massive amounts of data would have been impossible before AI (artificial intelligence). “These are the numbers behind genetic sequencing and artificial intelligence. To a degree never before possible, they give us power to understanding a pandemic even as it races to kill us.[i] ”
Recently when I was researching BA trends in the world of AI, I read several articles on the use of AI and quantum computing in healthcare, in particular drug development.[ii] One article even suggested that it would help predict pandemics.[iii] At the time that notion seemed far-fetched. Today AI is being used expensively to predict and try to stop the pandemic. Recent articles have cited numerous examples of the use of AI, HPC (high performance computing), and quantum computing[iv] and how they can help during this difficult time. Here are just a few ways these technologies are being used:
- Diagnosis, treatment, and vaccines to distinguish between pneumonia and COVID-19
- Identification of COVID-19 antibody candidates
- Examination of cell membranes to determine how the virus proliferates
- Predicting the path of disease spread
- Distribution of vaccines and personal protective equipment (PPE) like masks and ventilators, as well as medical personnel based on what is and will be needed and where
- Analysis of social media to predict where outbreaks will occur
- Analysis of geographic anomalies in body temperatures to predict the path of the virus
- Prioritizing patients with more urgent need and recommending individualized treatments
This is interesting and important information that will certainly help with future waves of this particular coronavirus, as well as future viruses. But what does this have to do with our world of projects? Plenty.
Need for accurate data
Since organizations began undertaking large AI initiatives, they have realized the importance of having not only massively large amounts of data, but of having that data be accurate. Here’s why. At its most fundamental level, machines use historical data (big, big data) to learn and improve. As new data becomes available, machines categorize it, learn more, improve, and make better and better predictions. Their algorithms depend on good data. If the data is unreliable, the outcomes will be less than useful and perhaps even harmful.
BAs have always been involved in helping organizations ensure that their data is accurate. They help organizations develop a business case for deciding whether or not to undertake the effort to cleanse the data, often a large and expensive undertaking. BAs can explain that while the cost to cleanse data is high, the risk of not doing so is also high. BAs can point to study after study of organizations that use inaccurate data and the disappointing results they have gotten from their AI efforts.
In addition, BAs can help organizations with this cleansing effort by doing such things as:
- Analyzing data to determine how much needs to be cleansed
- Developing cleansing implementation plans
- Facilitating conversations to help resolve the conflict related to, for example, who owns the data, where the data comes from, and which source of the data is the one that should be used
- Analyzing the results of AI simulations and questioning anomalies
BA as Data Translators
More and more organizations are recognizing the need for data translators. This is a perfect role for BAs who have always been good at translating technical complexities into language a variety of business stakeholders can understand. This is harder than it sounds. Stakeholders usually have their own language, acronyms, and idioms—think hospitals, insurance companies, doctors, nurses, medical staff, patients, first responders, and the community at large. BAs can help organizations figure out a way to communicate AI results to various stakeholders so that the communication is understandable and relevant to their specific needs.
In addition, BAs can help by:
- Looking at AI results and identifying trends, explaining the impacts of those trends, and explaining the importance of the assumptions used to come up with the results (see below for more on assumptions).
- Helping ensure that the results of simulations and predictions make sense to various stakeholder groups.
- Helping ensure that the results are visually understandable. Ill-defined and confusing charts and graphs are not useful for decision-making.
- Helping shape AI strategies as well as helping to implement them.
Strategic, experienced, and well-informed BAs can be consultants to their organizations, resolve conflicts between stakeholder groups, and balance competing needs among them.
Correcting faulty assumptions
BAs are good at questioning assumptions. We know that every assumption is a risk and that we need to be aware of and document them, so that when they change, we can easily change our plans. Take the simulations used in dealing with Covid-19. Since no one simulation provides enough information, multiple ones are being used. Some are based on assumptions about social distancing, although that is just one. Here are just a few more examples of assumptions that have been recently used in predicting the pandemic outcomes[v]:
- Everyone has the same chance of catching the virus from an infected person because the population is perfectly and evenly mixed, and that people with the disease are all equally infectious until they die or recover.
- Dividing the population above into smaller groups by age, gender, health status, employment, and so forth.
- The percentage of infected people vs those who died
- The number of days before an asymptomatic but infected person spreads the virus to others.
- That the data being used was accurate. For example, most simulations used data that came from China, which turned out to be inaccurate and therefore skewed the results.
As Dr. Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, said, “What we do is that every time we get more data, you feed it back in and relook at the model. Is the model really telling you what is actually going on?. . . Models are as only as good as the assumptions you put into them, and as we get more data, then you put it in and [the results] might change.”
BAs can help ensure that stakeholders understand the assumptions that are used and the effect of assumptions on the results. They can encourage running many simulations and looking at ranges of predictions based on different assumptions and communicating which assumptions were used for which results. And as tempting as it sometimes is, BAs can point out the risk of letting AI make decisions based on the results when assumptions are used.
[iii] Matt Swayne, March 4, 2020, How Quantum Computers an Be Used to Thwart a Future Pandemic, https://thequantumdaily.com/2020/03/04/how-quantum-computers-could/HPC Wire, March 12, 2020
[iv] https://www.hpcwire.com/2020/03/12/global-supercomputing-is-mobilizing-against-covid-19/, Global Supercomputer is Mobilizing Against Covid-19
[v] The Simulations Driving the World’s Response to Covid-19, David Adam, Nature, April 2, 2020, https://www.nature.com/articles/d41586-020-01003-6