S-115 Course

A Learning Guide to "Voices in the Code"

Developing Panoramic Thinking for an AI-Integrated World

Welcome to a guided exploration of David G. Robinson’s Voices in the Code. This is more than a book about medicine or technology; it is a profound case study in how human values, social structures, and technical rules collide to make life-and-death decisions.

Algorithms are already shaping your world, from the videos you see online to the news you read and, increasingly, to opportunities you may get for school or work. As the STAT S-115 course framework highlights, the specific code and models of today will be obsolete in a few years. What will not be obsolete is the ability to think critically about any such system, regardless of its technical architecture.

This guide is designed to help you practice panoramic thinking: the ability to analyze a complex system from multiple perspectives at once. Using the U.S. kidney allocation algorithm as our case study, you will learn to ask questions and integrate insights from four distinct but interconnected domains:

The Technical Lens: How does it work? What are its rules and inputs?
The Ethical Lens: Is it fair? What moral trade-offs are being made?
The Social & Historical Lens: Who wins and who loses? What is the human context?
The Policy Lens: Who makes the rules? How is the system governed and held accountable?

By the end of this journey, you won't just understand how one critical algorithm works; you will have a powerful, transferable framework for analyzing any high-stakes system you encounter in your life and career.

Chapter 1: The Human Values That Hide in Algorithms

The Big Picture: Robinson opens by establishing the book’s central argument: algorithms are not neutral, objective tools. They are "the instruments of our values". This chapter presents six brief case studies—from automated hiring to welfare fraud detection—to show that ethical choices are often "buried under a mountain of technical detail". The goal is to demonstrate that the difficult moral and political work of governing these systems is both necessary and, all too often, absent.

Key Concepts:

Algorithm: Robinson uses the layperson’s definition: "rules carried out by software, on a computer". This is distinct from the broader computer science definition which could include a person doing long division.
Moral Anesthetic: The idea that turning a difficult ethical problem into a quantitative one can make it "seem neutral and objective" , allowing us to make hard choices "without seeming to decide".
High-Stakes Algorithm: A system that makes life-altering decisions, often about vulnerable people who lack wealth and social capital.

Panoramic Thinking Practice:

Technical Lens: Robinson notes that programmers "inevitably alter established rules when embedding them into [software] code". Why might this be the case? Think about the difference between a vague legal principle (like "fairness") and the precise instructions a computer needs.
Ethical Lens: Consider the "NarxScore" used to predict opioid abuse. A woman was denied care because of painkillers prescribed for her dogs. The algorithm correctly identified a pattern (many prescriptions), but was its conclusion fair? What does this tell you about the limits of using data as a proxy for human behavior?
Social Lens: Robinson observes that algorithms are frequently used to make decisions about people "at the bottom or the edges" of society, like those seeking welfare or appearing in criminal court. Why do you think powerful institutions are so eager to apply algorithms to these populations specifically?
Policy Lens: For the New York City school screening algorithms, the city initially claimed it couldn't provide a full report on admissions criteria because they were not "centralized". What does this tell you about the challenges of governing a complex, decentralized system? If you were a policymaker, what would be your first step to create accountability?

You're in Charge:

Imagine you are on the "automated decision systems task force" for your city. Your first task is to decide which city algorithms count as "high-stakes" and require special oversight. Based on the examples in Chapter 1 (hiring, insurance, medicine, school admissions, criminal justice, welfare), what are the top three criteria you would use to make this determination?

Chapter 3: A Field of Life and Death

(Note: We skip the detailed theoretical discussion in Chapter 2 to focus on the core narrative).

The Big Picture: This chapter tells the origin story of kidney medicine. It's a dramatic history of brilliant innovations that solved one problem only to create a new, deeply moral one. Dr. Belding Scribner's Teflon shunt transformed kidney failure from a terminal diagnosis into a chronic condition, but it immediately created a terrible scarcity problem: who gets to use the few available life-saving machines?

Anecdote to Anchor On: The "Seattle God Committee"

Faced with more patients than dialysis slots, Scribner and his colleagues did something extraordinary: they refused to make the choice alone. They created a committee of anonymous laypeople—a lawyer, a minister, a housewife, a state official, etc.—to decide "Who Lives, Who Dies". This committee made explicit value judgments, considering factors like a person's "character and moral strength," their number of dependents, and their potential to return to work.

Panoramic Thinking Practice:

Technical Lens: Dr. Scribner's innovation wasn't a complex computer; it was a simple tube made of Teflon that prevented blood from clotting. How does this "low-tech" example reinforce Robinson's argument that the impact of a technology, not its complexity, is what creates high-stakes choices?
Ethical Lens: The Seattle committee was criticized for applying "middle-class suburban value system[s]". Yet, Robinson seems to find something to "admire" in their approach. What is the ethical argument for making these choices explicitly, even if the criteria are debatable, versus using a seemingly "objective" rule like first-come, first-served?
Social & Historical Lens: The federal government’s 1972 decision to have Medicare pay for all dialysis treatments was a landmark social policy. How did this act of public generosity paradoxically create the "mortal waiting room" for organs by keeping thousands of patients alive long enough to need a transplant? This is a classic example of a solution to one problem creating another.
Policy Lens: Initially, the choice of who got dialysis was made by doctors and a local committee. By 1972, it became a matter of federal law. How does this shift from a local, ad-hoc process to a national, formalized entitlement change the nature of the decision-making?

Thought Experiment:

You are a member of the 1962 Seattle committee. Your two final candidates for one dialysis slot are a 35-year-old brilliant scientist with no children who is working on cancer research, and a 45-year-old mother of six who works part-time as a librarian. How would you even begin to make this choice? What additional information would you want? What does this exercise tell you about the appeal of using seemingly neutral rules to avoid such a decision?

Interlude & Chapter 4: An Algorithm in Focus

The Big Picture: This is the heart of the book. We see the entire decade-long process of redesigning the kidney allocation algorithm, from a bold initial proposal to a messy, compromised, but ultimately successful implementation. It is a masterclass in how participation, forecasting, and shared infrastructure work in the real world.

Key Concepts:

LYFT (Life Years From Transplant): The first major proposal. It was a complex model designed to be purely utilitarian—that is, to maximize the total number of life-years gained from all available kidneys. It gave higher priority to younger, healthier patients.
KAS (Kidney Allocation System): The final algorithm that was implemented in 2014. It was a compromise, balancing the goal of maximizing life-years ("utility") with goals of fairness and access for older and minority patients ("equity").
Equity vs. Utility: The central ethical trade-off of the entire debate. Should we give an organ to the person who will live the longest with it (utility), or should we ensure that everyone, regardless of their age or health, gets a fair chance (equity)?
DSAs (Donation Service Areas): The arbitrary geographic zones that historically determined organ distribution. Robinson shows how this "off-stage" factor, which was not part of the KAS redesign, had a huge impact on fairness, but was too politically difficult to address initially.

Panoramic Thinking Practice:

Technical Lens: The LYFT proposal was abandoned partly because it was too complex for patients and even many doctors to understand. The final KAS used a simpler "longevity matching" system for the top 20% of kidneys and recipients. What does this tell you about the relationship between an algorithm's technical complexity and its social acceptability?
Ethical Lens: At the 2007 Dallas forum, patient Clive Grawe argued that the LYFT model would "penalize" him for living a healthy life that delayed his need for a transplant into his fifties. This personal testimony was a key reason LYFT failed. Why was his moral argument—based on fairness and incentives—so powerful against the utilitarian argument of maximizing life-years?
Social & Historical Lens: The final KAS included a crucial change: calculating a patient's waiting time from the day they started dialysis, not the day they were added to the list. This was a direct attempt to correct for the fact that minority patients were often referred to the waiting list later in their illness. How does this show the algorithm being modified to correct for injustices happening outside the allocation system itself?
Policy Lens: The redesign process took a full decade. For years, the committee was stuck because it couldn't get a clear ruling from the government on whether using age was discriminatory. Then, a series of lawsuits over lung and liver allocation forced the issue of geography, leading to the abolishment of DSAs. What does this saga teach you about the limits of a consensus-based, participatory process versus the power of courts and government mandates?

You're in Charge:

You are a data scientist at the Scientific Registry of Transplant Recipients (SRTR). The policy committee asks you to simulate two versions of a new algorithm.

Version A will save an estimated 10,000 total life-years, but it will reduce the transplant rate for patients over 65 by 30%.
Version B will save only 5,000 life-years, but will not change the transplant rate for any age group.

Your job is not to choose the policy, but to design the presentation that explains this trade-off to a public forum of patients, doctors, and ethicists. What charts, graphs, and explanatory text would you create to make the moral trade-off as clear as possible?

Chapter 5 & Conclusion: Ideas for Our Future

The Big Picture: Robinson extracts the core lessons from the transplant story, creating a blueprint for how we might govern other high-stakes algorithms. He argues that sharing the moral burden is possible, but it requires specific kinds of infrastructure and a willingness to engage in slow, messy, democratic processes.

Key Ideas for Your Panoramic Toolkit:

Algorithms Direct Our Moral Attention: The way an algorithm is designed focuses our debate on certain variables (like a patient's age) while leaving others (like geographic boundaries) off the table. The engineering itself frames the ethical discussion.
Participation Shapes Opinions, Gradually: The goal of public participation isn't just to "collect input." It's a "tumbler" that wears down sharp edges, builds shared understanding, and allows a community to change its mind and converge on a tolerable compromise.
Shared Understanding Needs Shared Infrastructure: Meaningful debate is only possible when everyone has access to trusted, independent analysis (like the SRTR's forecasts) and clear, plain-language explanations. This infrastructure is a public good that must be intentionally built and funded.
Quantification Can Be a Moral Anesthetic: Using numbers and scores can make a horrifying choice (who gets to live?) feel like a neutral, technical calculation. Robinson argues this is "often, but not always, a bad thing". Sometimes, it's a necessary "subterfuge" that allows society to function in the face of impossible trade-offs.
Knowledge and Participation Don’t Always Mean Power: Even the best-designed participatory process can be overruled by higher authorities like courts or legislatures. Governance is a multi-layered system, and these "venues of final appeal" are always part of the landscape.

Capstone Reflection: Applying the Panoramic Framework

You have now analyzed one of the most complex and high-stakes algorithms in the world. Your final task is to take the panoramic framework and apply it to an algorithm from your own life.

Choose one of the following:

Your TikTok or YouTube recommendation feed.
Your Spotify or Apple Music "For You" playlist generator.
A college's financial aid or admissions screening tool.
A "risk assessment" score used by a car insurance company.

Write a short analysis (500 words) using the four lenses:

Technical: What do you think the key inputs are for this algorithm? (e.g., watch time, likes, user demographics, zip code, etc.) What is it trying to predict or optimize for?
Ethical: What is the "utility" this algorithm is trying to maximize (e.g., user engagement, ad revenue, identifying successful students)? What "equity" considerations might it be ignoring (e.g., exposing users to diverse viewpoints, fairness to low-income applicants)? What are the hidden value judgments?
Social/Historical: Who are the winners and losers in this system? Does it benefit some groups over others? Does it change social behavior (e.g., how people create content, how they drive, how they study)?
Policy/Governance: Who is in charge of this algorithm? Is there any transparency into how it works? Is there an "SRTR" that audits its performance? Can you appeal a decision you think is unfair? What would a more accountable system look like?

By completing this exercise, you will have moved from understanding a single case study to possessing a durable intellectual tool. The technology will constantly change, but your ability to think panoramically about its impact on the world will remain an essential skill for the rest of your life.