Managing Complexity

Navigating the rules of the health insurance can be complex.

A majority of insured adults (58%) say they have experienced a problem using their health insurance in the past 12 months – such as denied claims, provider network problems, and pre-authorization problems.

This is from a recent Kaiser Family Foundation (KFF) survey of 3,605 U.S. adults with health insurance. It also found:

Nearly half of insured adults who had insurance problems were unable to satisfactorily resolve them, with some reporting serious consequences.

Yet the study found most viewed their health insurance favorably. Those with good health rated it higher than those reporting poor health.

 

Ben Dickson sheds light on what is behind ChatGPT and Microsoft’s Bing Chatbot named Sydney. These large language models (LLMs) will likely find their way into our normal routines. It is important for us to understand their strengths and weaknesses before is gets your coffee order wrong and eloquently explains to you why it was right.

His well-written post “To understand language models, we must separate “language” from “thought” describes what these LLMs, created through machine learning of massive data sets, do well and what they struggle with. He cites a recent paper Dissociating language and thought in large language models: a cognitive perspective that found:

LLMs show impressive (although imperfect) performance on tasks requiring formal linguistic competence, but fail on many tests requiring functional competence.

Leadership can be a lonely place. Especially during the Covid-19 pandemic if you are restaurant owner, school superintendent, healthcare leader, or wedding planner. It is impossible to process the complexity of the many knowns, unknowns and uncertainties to forecast the future and satisfy the impacted people. Yet, that is what we ask them to do.

As an example, here are 15 Covid-19 forecasts healthcare leaders must get right. See more.

The global pandemic has forced a curriculum change to understanding the complexity of our modern life. We were auto enrolled in immunology, epidemiology and forecasting lessons that are prerequisites to make sense of therapeutics, vaccines, and virus transmission based on human behavior. This understanding will help us forecast R value infection rates, vaccine timelines, eventual herd immunity and a new date for the family reunion.

The brilliant and talented Atlantic writer Ed Yong provides a ‘must read’ lesson on immunology and COVID-19, The Pandemic’s Biggest Mystery Is Our Own Immune System. Ed is a great explainer. If you like this article, I would strongly recommend you read I Contain Multitudes: The Microbes Within Us and a Grander View of Life. He explains the complexity of science in a fun and easy to read way. It is one of my favorite books.

The past six months has provided us an education in epidemiology, virology, and human behavior. While we didn’t sign up for these lessons, they became prerequisites for making sense of living during a pandemic and predicting what’s next. How will our kids be educated? Will elections be safe and fair? Will the newly unemployed be hired? Will the entertainment businesses survive?

We are seeing signs of coronavirus outbreaks slowing in Arizona, Texas and Florida. This is likely the result of human behavior changes in staying home, wearing masks, hand washing, and social distancing. We are also learning valuable insight to how our immune systems combats the coronavirus. This new insight may not impact the fall, though it may have a major impact on 2021.

Our National Institute of Health (NIH) Director Dr. Francis Collins highlighted this week new insight on how our T cells fight COVID-19. While the antibodies produced by our immune systems get most of the media attention related to vaccines, therapeutics and testing, the human T lymphocyte (T cell) needs to be added to our vocabulary and discussions.

The findings of a new study reported in Nature, suggests these T cells may protect some people newly infected with SARS-CoV-2 by remembering past encounters with other human coronaviruses. Four of the six known coronaviruses are responsible for the common cold. This might explain why some people get very sick while others are asymptomatic.

This study indicates we need to add T cells to the general publics epidemiology and virology curriculum. T cells may soon become part of our testing, therapeutics and vaccines discussions.

When it comes to managing the complexity of living during a pandemic,
developing predictions based on our understanding of many impacting elements is essential. These predictions will become the foundation for our strategies to new manage the complexity of 2021.

Image: Scanning electron micrograph of a human T lymphocyte (T cell) from a healthy donor’s immune system from National Institute of Allergy and Infectious Diseases/NIH.

When we decide we want to improve our health, transform our business or launch a new product, we develop a plan and execute it. As described in Why We Struggle with Complexity, we often apply logic and analogies of similar circumstances to make sense of these complex challenges and to determine what to do. We know logic (if this, then that) and analogy (like that, except this) decision making can fail miserably with complexity.

Logic becomes ineffective when there are too many variables or when variables are unknown or unpredictable. Analogies becomes ineffective when the context, people and variables are too different.  If you failed to achieve a complex goal after hard work and developing the required knowledge, it is likely due to oversimplifying complexity with logic and analogy. You were likely impacted by one or more of these factors:  

• Too many impacting variables for logic and analogy reasoning methods
• Unknown impacting variables that are difficult to identify
• Uncertain impacting variables difficult to predict how they will act, react, or interact
• Unpredictable outcomes due to change and the combination of the variables

Managing complexity is not new to scientists, inventors, scholars, and system engineers. They learned they need to perfect the process of:

• Understanding and predicting the complex system
• developing strategies that achieve the mission while addressing the complex system
• iterations of adjusting the strategy and model of the complex system

With complexity, a definitive answer based on logic or analogy is not possible. The logic and analogy reasoning that we were taught in school, is no match for too many, unknown, uncertain, and unpredictable variables. The best strategy comes from many iterations until the mission is achieved. Here is a closer look into how they manage complexity.

Scientists such as biologists, ecologists, meteorologists, psychologists, and sociologist model how numerous elements (variables) will act, react, and interact (complex system). Meteorologist use these predictive models forecast rain, wind, temperature and even hurricanes. When their weather predictions miss their mark, they make incremental adjustments to their models of the complex system, change and strategies. This results in a continuous cycle of improvement in forecasting.

Inventor Thomas Edison built a model for managing the complexity of invention. His model included a Menlo Park research laboratory, investor funding and teams of people with diverse domain knowledge. He used this invention framework to make over 3,000 incremental adjustments to his lightbulb creation. He had to address the complexity of finding the right balance of low production costs, durable filaments, brightness, and low energy consumption. While Edison did not invent the lightbulb, he did invent a bulb that fit within his “seven-point program” to be competitive with gaslighting. The seven points included “the parallel electric circuit; the durable, high-resistance light; the improved electrical generator; underground copper wiring; devices to keep the electrical current constant no matter the location of the bulb; safety fuses, insulating materials and usage meters; and lighting fixtures with switches to turn them on and off (1). Since electric power distribution did not exist, Edison had to manufacturer most of the devices. His first light bulb became commercially available in 1879 and he introduced the first electrical power plant and distribution system in New York in 1882 to create demand for his innovation.

Hedge fund firm Bridgewater Associates built a model to predict financial markets and to test their investment theses. To ensure as many plausible viewpoints are considered about the model’s impacting elements (complex system) and investment thesis (strategy), company founder Ray Dalio instituted a corporate framework of meritocracy. Bridgewater hiring strategy ensures a wide diversity of domain knowledge and meritocracy strategy solicits the best viewpoints including from the new hires just out of college. They document the success of employee theories, inputs, and persuasiveness to understand their decision-making process. They then make incremental adjustments to their predictive models, decision making process and investment theses based on what is learned from their multi-billion dollars of investments. Bridgewater used their predictive modeling and culture of meritocracy to grow to manage $150 billion of capital and one of a few organizations correctly predicting the 2008 Great Recession. 

The system engineers of Amazon, Apple, Facebook, and Google show us how to thrive with managing complexity within frameworks with thousands, or millions of incremental improvements. It would be hard to imagine how we managed our lives with the early versions of Amazon (1994), Apple iPhone (2007), Facebook (2004) and Google (1998). These companies have become industry leaders in managing complexity within their frameworks with the use of Artificial Intelligence based on data science and deep learning algorithms. Google search analyzes “short clicks” versus their “long clicks” to continuously make incremental improvements to their solution. If users click on the first search result and does not come back (“short clicks”), the were likely satisfied. If users click on the first search result, spends time there (“long click”), then clicks and become satisfied with the second result, the deep learning neural network may make slight adjustments to the weighting of some of the hundreds of search signals and sub-signals.

Managing complexity is not something most people are taught in school, that is unless they enroll in advanced college degree programs.  Master’s and post-doctorate degree programs study complex systems, such as biology, ecology, economics, meteorology, political science, psychology, or sociology.  While these degree programs teach the complexity of their subject domains, they have little time to study the fundamentals of complex systems or how to address the complexity that fills our lives.

To manage complexity, scientists, inventors, scholars, and system engineers use frameworks to understand their complex systems and predictive modeling of their strategies. The use iterations of incremental improvements to their frameworks and predictive models as they learn. They determine the most effective strategies (forecasts, innovations, theories, and products) through iterations of trial and error, learning and adjusting their frameworks and models until they get it right.

To achieve complex goals, we can apply some of the same complex system and iteration thinking principles used by experts. We learned logic & analogy methods in school in a formal way. We learned complex system and iteration methods in life in an informal way while achieving our complex goals.  We accomplish complex goals by making sense of the complex system, developing an effective strategy and iterations of “learning by doing” until we achieve our mission.

Logic & Analogy	Complex System & Iteration
Few variables	Many variables
Known variables	Known & unknown variables
Certain variables	Certain & uncertain variables
Rules & formulas	Frameworks and models

To be successful achieving complex goals to improve our health, transform our businesses or launching a new product, we need to be efficient and effective assessing the situation (complex system) and developing a plan (strategy) to achieve it. Complex system and iteration thinking can help get you there. To create frameworks of complex goals (complex systems), strategies and to manage iteration like experts, it requires understanding the eight essentials to managing complexity.

(1) Josephson, Matthew, Edison: A Biography. McGraw-Hill, 1959, p229

For the first time since 1962 and 1963, the United States life expectancy at birth has declined two years in a row.  This has been mostly attributed to higher death rates among young and middle-aged Americans with a 21 percent increase in fatal drug overdoses and the death rate doubling from synthetic opioids like fentanyl.  This overshadows the progress made on reducing deaths from seven of the ten leading causes of death (heart disease, cancer, chronic lower respiratory disease, stroke, diabetes, influenza and pneumonia, and kidney disease) over the past year. The three of the ten leading causes of death that contributed to the decline in life expectancy were unintentional injuries, Alzheimer’s disease, and suicides.

We have added 38 years to life expectancy over the past 150 years, though our progress has seemed to hit a wall.  We learned how to address and eradicate infectious diseases with immunizations, reduce deaths from infections with the discovery of antibiotics in the 1930s, reduce heart attacks through education, medicines and treatments, reduce cancer deaths with smoking awareness and inventing chemotherapy and radiation treatments, as well as improving health outcomes by identifying maladies earlier with the advances in medical imaging.

Unless we invent something as profound as antibiotics or immunizations, it is unlikely we will make much progress without a clear mission and new approach. Here are the areas that need improvement:

1. Managing Complexity – To get better with improving the top ten leading causes of death.

Complexity – the unpredictability of many impacting elements (people, communities and realities) that operate, interact and react in both certain and uncertain ways.

We must go beyond managing the complexity of medical treatments. We must address the 80-85% of health determinants (impacting elements) that contribute to health outcomes that are outside the health care system. These impacting elements include behavior, health literacy, environment, socioeconomic and other social determinants. To get better with managing the complexity of health, we will need to improve our understanding of how these thousands of certain and uncertain elements are impacting our individual health.

2. Mission – Develop clearly defined missions.

Our mission probably should be to improve health span (years without functional limitations) rather than live span. Who owns improving the health span mission? This is obviously a mission beyond any one organization. Yet we can agree it is a worthy mission. Therefore, we need to link the many missions that could contribute to achieving this mission.  The improvement in seven of the leading causes of death (heart disease, cancer, chronic lower respiratory disease, stroke, diabetes, influenza and pneumonia, and kidney disease) was due to many organizations with missions to improve health outcomes without anyone owning the entire mission. By linking the missions together, we can identify areas of impacting elements (health determinants) that are not being addressed.

3. Strategies – that make progress toward the mission and that address the impacting elements

An example of this is the four strategies offered by Dr. Joshua Sharfstein to address the increasing death rate from opioids in an article in Journal of the American Medical Association.

Addiction Treatment – The use of the opioid agonists methadone and buprenorphine reduces overdose, illicit drug use, crime, and transmission of infectious diseases.

Criminal Justice – It is now recognized by many across the political spectrum that the arrest and jailing of millions of Americans for their addiction has complicated efforts to address the opioid epidemic.

Health Care System – There is now broad understanding that the overprescribing of opioids has contributed to today’s opioid epidemic.

Looking to Evidence – On opioids, it can sometimes seem that there are 3 bad ideas for every good one. Public officials have supported limiting the number of naloxone resuscitations and afterwards letting people die, requiring drug testing before enrolling in Medicaid, and launching stigmatizing public relations campaigns that can reduce the chance people will seek treatment.

Managing these strategies to reduce the death rate from opioids will not be easy. These strategies could benefit, as well as the ten leading causes of death, from an overarching strategy to address the impacting elements (health determinants) that are common across each of the leading causes of death.

4. Solutions – interventions that get us closer to our mission and address the impacting elements

It would require understanding how to cost-effectively manage the complexity of the many interacting and uncertain elements that impact health outcomes.  It may require thousands of different medical and non-medical interventions that can be measured for cost-effectiveness. If the solutions are not cost effective, they cannot be sustained.

5. Measures – to measure progress

This includes measuring the progress toward achieving the mission and the understanding of the impacting elements. We must also measure the performance of strategies and solutions to determine what works.

The approacch to measure could be similar to what Google does with its search algorithm, Facebook does with engagement improvement and Amazon does with moving merchandise. They have created frameworks to manage complexity by using data to provide insight on opportunities to drive incremental improvements.

While genetics plays into many of our chronic diseases that limit our function (health span) and contributes to early deaths (life span), we also need to understand how the non-biological factors (impacting elements) impact and influence gene regulatory functions.  This will not be easy. It will be a worthy journey rather than a destination.

There is no doubt artificial intelligence will help us in the future with managing the complexity of improving health span. Yet, for artificial intelligence to help us get over the complexity wall, we will need a framework and massive amounts of data for it to be effective like it is with Google, Facebook and Amazon. To create this framework and data, we need to address the five areas listed above as soon as possible. This will provide insight on how to make incremental improvements every year to health span and life expectancy like we have done for most of the past 150 years.

Why do good ideas fail? Why don’t business strategies achieve their mission? Why doesn’t expert career advice work for us? Why hasn’t proven weight-loss methods delivered our desired weight? Why wasn’t knowledge and hard work enough for success? The surprising answer to these questions is likely the same. It’s our struggle with complexity.

To achieve something meaningful, we must manage the uncertainty of complexity, yet we’re not taught that in school. Complexity involves many interacting certain and uncertain variables, that impact our health, intimate relationships, transforming organizations or finding cancer cures. We are taught in school to use known and predictable variables, logic and formulas to find answers. We learned through literature, history and science how people navigated complex challenges and to identify relevant similarities to infer analogies to help guide us though our challenges.

While these approaches work in school, they often fail miserably when applied to the complexity of transforming our health, careers, businesses and nation.

Complexity – the unpredictability of many impacting variables (known and unknown) operating, reacting and interacting in both certain and uncertain ways.

Scientists, inventors, scholars and system engineers have studied complexity for centuries to make sense out of biology, ecology, economics, politics, psychology, social science, weather and creating new products. They have long known that managing complexity requires complex system frameworks to understand the impacting variables and the use of an iterative process of incremental improvements to be successful.

When faced with the uncertainty of complex decisions, experiments by psychologists Amos Tversky and Nobel Prize winning Daniel Kahneman demonstrated that we use mental shortcuts, cognitive biases and bypass facts. They found we use logic based on a small subset of the impacting variables while ignoring many important factors. They observed how our decision are influenced by our cognitive biases based on analogies with relevant similarities.

While the logic and analogy thinking we learned in school is effective with most everyday decisions, its effectiveness diminishes with each increase in complexity. It contributes to our struggle with complexity when there are too many variables, unknown variables and unpredictable variables. While logic and analogy thinking has improved our standard of living, it has limits in getting us to the next level in our modern day of expanding complexity. To manage complexity, we can learn from how scientists, inventors, scholars and system engineers do this every day.

When people break into banks, they don’t tend to deposit money. Which explains why JPMorgan Chase and Wells Fargo spend billions of dollars each year on artificial intelligence, physical security and cybersecurity to prevent financial and identity theft.

When Cambridge Analytica compromised 87 million Facebook users’ data, they deposited millions of dollars into Facebook’s bank account. Which may explain why Facebooks’ massive, world-class team of Artificial Intelligence professionals in Building 20 were not working to prevent bad actors from purchasing ads and stimulating users emotions with sensational material (a.k.a. click-bait).

Cambridge Analytica claims they have 5,000 data points on American voters. They leveraged Facebook’s artificial intelligence powered capability that predicts how each of their two billion users will behave, think and purchase. Yes, if you are a Facebook user, they may be able to predict your future behavior better than you. Advertisers have found this insight to be so valuable that they helped drive Facebooks net income to $15 Billion in 2017.

What is Artificial Intelligence?

Artificial Intelligence is the machine’s (computer system) ability to perform tasks that normally require human intelligence and the capability (machine learning) to improve its performance without humans explaining how to do it.

The artificial intelligence buzz, or internet click bait, has promoted fear that artificial intelligence may become the supreme ruler of mankind and that it will one day eliminate your job. Word travels fast when Elon Musk states AI is more dangerous than nuclear weapons and when McKinsey predicts that 73 million jobs in the U.S. will be eliminated by 2030. There are few views, shares and likes for another McKinsey study which predicted AI will drive as much as $5.8 trillion of new economic value across nineteen industries after they analyzed four hundred AI use cases.

Most AI Resources Focus on What Humans Can’t Do

The popular stories that go viral include AI beating champions of chess, Jeopardy and Go, yet quietly Facebook’s AI experts tune their machine learning models to predict how each of their 2 billion users will think, act and buy.  We are inundated with AI-powered autonomous vehicles stories that forecast the elimination of millions of driving jobs, while Goggle’s AI team behind the scene are tuning their machine learning models to you, your context and what you my buy online. We are no longer surprised by stories of how AI is able to detect cancer on medical images as well as radiologists, yet we don’t hear how Bridgewater Associates’ AI team is effectively turning their AI frameworks to predict financial markets for their investors of over $150B.

We mostly hear about the small fraction of AI investment going to what humans can do, such as play games, drive cars and diagnose cancer. We hear very little about the objectives of most AI investments, which is to get machines to perform that haven’t ever been effectively done by humans.  Even if humans can predict the human behavior of a single person from Cambridge Analytica’s 5,000 data points, could humans effectively do that with 87 million users? Could humans effectively make sense of billions of data points processed by Google’s and Bridgewater’s machine learning models that predict what people want and the financial markets?

AI Investment is Going to Managing Complexity

Complexity is the unpredictability of many impacting elements (people and realities) operating, interacting and reacting in both certain and uncertain ways. To achieve a complex challenge, we must manage the uncertainty of complexity, yet we’re not taught that in school. Complexity involves many certain and uncertain elements that interact, and get in the way of our health, relationships, career goals, transforming organizations or finding cancer cures. Yet, we are taught to use certain known elements and formulas to find answers, which doesn’t help with complexity. While proven formulas and calculations may work in school, they fail to address the complexity of predicting human behavior, what people want and the financial markets.

This explains why Facebook, Google and Bridgewater abandoned their predictive algorithms and replaced them with Artificial Intelligence fed by machine learning models. Facebook CEO Mark Zuckerberg sees AI as the only way to address many complex challenges.

Zuckerberg referred to AI technology more than 30 times during ten hours of questioning from congressional lawmakers Tuesday and Wednesday, saying that it would one day be smart, sophisticated and eagle-eyed enough to fight against a vast variety of platform-spoiling misbehavior, including fake news, hate speech, discriminatory ads and terrorist propaganda. (Washington Post)

Mark Zuckerberg cited AI as the solution for many of the most complex challenges.

Moderating hate speech? AI will fix it. Terrorist content and recruitment? AI again. Fake accounts? AI. Russian misinformation? AI. Racially discriminatory ads? AI. Security? AI. (The Verge)

How Does Facebook Manage Complexity with AI? 

The AI technology of Facebook, Google and Bridgewater have operated in the shadows behind protective proprietary walls while driving massive shareholder value. For Facebook, the shadows began to fade when stories emerged about bad actors using Facebook during Brexit and Presidential elections. Then came the Cambridge Analytica story, which resulted in Facebook losing $35B in shareholder value. Then details of the activities inside Facebook’s Building 20 started to emerge.

Facebook uses an artificial intelligence-powered prediction engine called “FBLearner Flow” that self-improves by inputs from its several machine-learning models. The machine learning models process billions of data points from its two billion users’ activity to predict thousands of things including: users in a photograph, what people want in their data feed and what is likely to be spam. Inside documents state that the AI system can predict your behaviors.

Rather than focus on replacing humans, Facebook uses AI to augment their human engineers. “We’re able to do things that we have not able to do before,” he says Facebook Chief technology officer Mike Schroepfer.

Humans can’t cost-effectively interpret massive data sets and make sense of thousands of certain and uncertain elements, that interact and react in unpredictable ways.  Software developers can’t create computer algorithms for each Facebook user nor can they develop algorithms that can interpret how thousands of certain and uncertain elements, will interact and react.  Artificial Intelligence, fed with machine learning models, are beginning to do what humans or computer algorithms can’t do to. They interpret what we want in our Facebook feeds, what we are searching for with Google and what the financial markets will do.

Lesson From Facebook

1. It’s What Humans Can’t Do – AI is not simply cost-effective automation that replaces what humans can do, it is a way to address complexity in way that wasn’t possible before AI.
2. Algorithms Can’t Do It – Predictive algorithms are limited and become less effective as the number of data elements grow. Facebook, Google and Bridgewater moved to machine learning models that can improve their performance without humans programming how to do it.
3. You’re Missing Data Elements – You will need data elements that you don’t have today. Facebook and Google capture data outside their applications to help their machine learning models. If you need more on this, asks members of Congress who were briefed by Mark Zuckerberg on internet cookies.
4. You Need More Data – Machine Learning requires massive amount of data to be effective. Most AI professionals work for companies that have massive data sets like Amazon, Netflix, and Apple.
5. You Better Start Now – Machine Learning Models take many years to develop. They need to go through many iterations to get the data elements, their associates and weighting right to be effective. It is not an algorithm that can be written, it could take 3 to 5 years of tuning machine learning models before they may be effective.
6. You Need to Know All The Elements That Impact Your Mission – When Facebook, Google, and Bridgewater didn’t predict human behavior or financial markets correctly, they found they were missing elements that impact human behavior or financial markets. You need to invest time and resources upfront to identify all the impacting elements or your AI system could drive poor business decisions.
7. You Are Never Done. When managing complexity, you need substantial resources to continue working to address change and the unpredictable elements. The machine learning models will need to be adjusted based on the thousands of changing realities and new insight.
8. Protect Your Data. It costs almost nothing to share digital data. It can cost $35B in shareholder value if your data ends up in the hands of organizations that use the data like Cambridge Analytica.
9. Add Digital Product Miss-use to Your Risk Mitigation Program. Don’t wait until your company is trending on Twitter before using AI to address mis-use of your digital systems like Facebook.
10. Artificial Intelligence Professionals Are Very Expensive. If you want to hire AI resources, you will need to compete with Google, Facebook, Apple, Amazon and Uber. Top AI talent are commanding one million dollar salaries even at non-profits.