Original article was published on Artificial Intelligence on Medium
Plotting a path out of the crisis
Crunching the numbers isn’t enough to transform data into action
During the SwissCognitive conference I gave online last week on putting AI in perspective, I studied the faces of several dozens in the audience to the right of my screen. Although their micros were muted during my talk, each figure offered me clues of their past experience, insights into how they live the present crisis, and glimpses of their perspectives of the future.
The social and economic consequences of the current Covid- 19 pandemic offer us a “Decisive Moment” where we can potentially choose to alter the course of history. Decision-making in times of uncertainty requires looking beyond the numbers to understand how the data has been framed, how we’ve defined the horizon, and in what conditions AI can augment human intelligence.
The Importance of Perspective
Human Perspective is the missing link between data and action
Like many confined to their home desks the last several weeks, I open my computer each morning to read the stats on the impact of the current pandemic. The International Monetary Fund projected last week that annual growth in Euro area will fall by 7.5 percent, while that in the United States will dip by 5.9 percent. The UN’s trade and development agency concluded that the coronavirus outbreak is likely to cost the global economy at least $1 trillion. with cutbacks equivalent to nearly 200 million full-time workers expected in the next three months alone. No matter which figures you choose, we need to ask how well they reflect reality, and more importantly how well they represent a call to action.
In the coming weeks we will have a choice of closing our eyes and resuming business as usual or seizing the opportunity to change the way we look at the world around us. How can we take better decisions given the uncertainty engendered by the current crisis to enlarge our perspective of alternative futures? Crunching the current numbers of cases, fatalities or jobs is at best only a start. If AI’s (machine learning, limited AI) ability to forecast the future depends of the relevance of past experience, seizing decisive moments relies on our ability to reframe what we see.
Why do decision-makers need to look beyond the data to develop a broader perspective of where alternative futures lie? If machine learning helps us identify patterns in the current crisis, making sense of the exceptions depends upon human sagacity. In the crisis of the Dark Ages almost 700 years ago, Leon Battista Alberti’s argued in De pittura that putting figures into perspective depends upon our frame of reference, how we draw the horizon between the present and the future, and nudging our audience to transform the data into action. Let’s explore what putting data into perspective implies today.
“The pertinence of the data depends upon the context in which it’s used”
What do the numbers of the current social and economic crisis really mean? Do the figures today provide us the means to predict what the future holds? Data is by its very nature an approximation of reality; the goal of Data Science here is not to improve the numbers but to nurture a call for individual and collective action. The precision, the granularity, and the accuracy of the figures contribute to our ability to analyze the past, describe the present, and predict the future. In stochastic decision environments the trust our audience has in the numbers is critical in both predicting and influencing the future.
Learning from the numbers begins with questioning how closely they represent the nature of the crisis we are facing. Alberti analyzed perspective from the observer’s point of view — what do the numbers represent for decision-makers and more importantly how can they impact behavior? Today semiologists study the signification of the numbers on three different levels. Semantics refers to the quality of the relationship between the figures and the events they are supposed to represent. Syntactics analyzes the coherence between the figures in each profession, organization, or market. Finally, pragmatics explores the relationships between the figures and the actions and reactions of the consumers, managers or stockholders who use them. Data is never objective, but facts have consequences.
In mid-May 2020, there were 4.4 million cases of Covid- 19 in the world, previsions of a drop in GDP of 7,5 % in Europe, and reports of unemployment rate of nearly 15 percent in the US. Yet the biggest consequence of the pandemic isn’t in the figures themselves, but in the risk, the ambiguity and the uncertainty around the numbers in plotting a path our of the crisis. When looking at the stats, we need to ask how they have been calculated, with what confidence intervals they been given, and how they represent a call for action. In our haste to meet the deadlines of news and reporting cycles, we zoom in on the numbers without examining the larger social and economic processes that have proven ill-adapted in addressing the crisis.
Value and values are never found in the data itself, but in how the data is framed
When we work with datasets, we are working with data that has been scraped together for a specific purpose. Alberti suggested that framing play an essential role in human perspective by helping us focus on what is important inside the frame in relation to that outside the frame. In analytics, frames are used to contextualize the data, suggest relationships between the records, and help understand how the data set can be leveraged in practice. Metadata are informational frames describe the underlying conditions which the data was captured, how it was acquired, its accuracy, and its method of compilation. Data taxonomies, or naming systems, are semantic frames used to classify data into categories and subcategories that reflect trade or business logic. Ontologies are models of knowledge, frames used to identify the proprieties and relationships between concepts and entities.
In the current crisis, the distinctions between morbidity and well-being, between employment and unemployment, and between action and inaction depend on how we frame the data. Metadata, taxonomies, and ontologies determine not only how we construct these concepts, but what types of conclusions we draw from each data set. These frames, rather than the data itself, help us decide which patterns, variables, and attributes are pertinent to address the questions under study, and which belong to other contexts and experiences. In both human and “automated” decision-making, framing operates at three specific levels: as a mindset of what the data is all about, as a set of beliefs of how problems should be resolved, and as a benchmark for detecting the outliers that defy traditional organizational practices.
The significance of the number of cases of Covid- 19 worldwide, the drop of GDP in Europe, and the unemployment rate in the US depends upon the frames we use to capture and qualify the data. While sanitary conditions, recessions, and social turmoil have at times provided the impetus for change, they are invariably symptomatic of the social and economic systems that produce them. Proponents of change management suggest that reframing our perceptions of normality, rather than crunching the numbers, is a necessary condition in helping communities and economies evolve. These “radical frames”, or sets of alternative values and interests, allow decision-makers to use the data to think differently about future. Do our definitions of well-being, prosperity, and employment still make sense today, or does the crisis invite us to reframe how we use the data to incite collective action today and in the foreseeable future?
Because the world isn’t flat, charting a course from the present to the future requires more than just connecting the dots
What does the future hold? In developing our perspective on our business and our communities, we fix a mental horizon that separates what we can see of the present from the challenges of the future just beyond our line of view. For Alberti, horizons corresponded to the boundaries of what human perception can either induce or deduce. Today, we form our perspective by defining points on the horizon using both heuristics and algorithms. We leverage either human and/or machine intelligence in perception, prediction, evaluation and insight depending on specific decision environments. The horizon is none-the-less clouded by our own perceptions of risk, prejudices, and our convictions. Looking beyond the horizon begins with recognizing the obstacles that obscure our point of view.
Unfortunately, life is not a straight line. In the real world we face uphill battles in addressing complicated, complex, and chaordic decision environments. Implicit bias in the data, the algorithms, and business logic limit our perceptions of what we see as well as what we do. Semantic drift has changed the very concepts of well-being, productivity, and value that structure the data we use. Perceptions of risk, uncertainty and ambiguity hinder our willingness to accept responsibility for transforming data into action. Our eagerness to invest in artificial intelligence blinds us to the importance of developing other dimensions of human intelligence (emotional, interpersonal, and even spiritual). Finally, our ethical frameworks concerning acceptable data practices condition how we use data to incite individual and collective action.
In systemic crises like the one we face today, apprehending the sanitary, economic, and social implications of the pandemic complicate decision-making. The pertinence of Covid testing, the appropriateness of social tracing, and the opportunity the balance between individual liberties and collective well-being are open questions. The decision environment is complex if not chaordic –for neither the causes nor the solutions are known. Our constructs of well-being, productivity and value are under siege, challenged by the very context in which we live. Algorithmic probability, even in context of deep leaning, will prove only as pertinent as our past experience. It is of little wonder that many of us may just staring at the numbers, even if it means missing the horizon completely.
The role of the oracle
Artificial intelligence by itself doesn’t solve problems, people do
In practice, machine learning produces new data points that can potentially reduce the risk, uncertainty, and ambiguity in decision-making. Data can be leveraged to improve our ability to take better decisions in one of three directions. Amplification refers to the clarification and application of a perspective in focusing on a set of events. Bridging involves the linkage of two or contingent but unconnected visions together in addressing an issue or problem. Extensions represent an effort to enlarge the boundaries of perspective to encompass the views, interests, or sentiments of a larger community. Yet, as the work of Peter Wason on confirmation bias demonstrated convincingly, data is often interpreted to simply support long-standing beliefs.
In what conditions can AI enrich human intelligence? The need to nudge, to incite, to motivate decision-makers to act on the data they see is a centuries-old concern. In the examples Alberti studied in pre-Renaissance painting, there was often a hand, a bird, or a symbol that literally breaks out of the frame to reach out to audience. These “oracles” symbolized a call to action, providing a line of vision that transgresses the frame to invite the spectator to look appropriate the artist’s vision of the future. This link between data and action is conspicuously absent today in our datasets, spreadsheets, and predictions, leaving many individuals and organizations ill-equipped to apply the data to their own organizational context and experience.
The challenge of the current pandemic is not flattening the curve but nudging our communities along a path leading out of the crisis. The data on the probable social and economic consequences of the crisis do not speak for themselves, but in the stories the we tell around the data, or in the skills and competencies of those responsible for producing results. The algorithms used in regression, classification, clustering and anomaly detection can provide sources of new knowledge, but developing human intelligence requires using the results to chart different paths out of the crisis than that we took coming in. Inciting individuals and organizations act decisively along these paths will depend upon developing human perspective of putting data in context, understanding how to frame the present, focusing on the horizon, and reframing the future.
Advancing human intelligence
Plotting a path out of the crisis won’t be done online, but in actively decisively in moving from where we are today to where we want to be tomorrow
As I finished my online discussion on AI’s role in enriching human intelligence, I took one last look at the cameos of the audience to the right of my screen. In the eyes, and in their questions, I had clues of their past experience, insight into how they lived the present, and a glimpse or two of how they were preparing life after the crisis. I knew that presenting the data on the sanitary social, economic crisis alone wouldn’t be enough to reveal the conditions in which machine and human intelligence could seize this decisive moment. I concluded, as we do in our training on the managerial challenges of AI, that plotting a path towards the future requires mastering the mechanics of putting the data in perspective.
The Business Analytics Institute (http://baieurope.com) offers executive training, coaching and mentoring to address the managerial challenges of artificial intelligence, digital ethics, and the development of “smart citizens”.
Lee Schlenker is the Principal in the BAI and Professor of Business Analytics and Community Management. His LinkedIn profile can be viewed at www.linkedin.com/in/leeschlenker. You can follow the BAI on Twitter at https://twitter.com/DSign4Analytics