Week 1: Introduction to the Course

DSAN 5450: Data Ethics and Policy
Spring 2025, Georgetown University

Jeff Jacobs

jj1088@georgetown.edu

2024-01-15

Who Am I? Why Is Georgetown Having Me Teach This?

\[ \DeclareMathOperator*{\argmax}{argmax} \DeclareMathOperator*{\argmin}{argmin} \newcommand{\bigexp}[1]{\exp\mkern-4mu\left[ #1 \right]} \newcommand{\bigexpect}[1]{\mathbb{E}\mkern-4mu \left[ #1 \right]} \newcommand{\definedas}{\overset{\text{defn}}{=}} \newcommand{\definedalign}{\overset{\phantom{\text{defn}}}{=}} \newcommand{\eqeventual}{\overset{\text{eventually}}{=}} \newcommand{\Err}{\text{Err}} \newcommand{\expect}[1]{\mathbb{E}[#1]} \newcommand{\expectsq}[1]{\mathbb{E}^2[#1]} \newcommand{\fw}[1]{\texttt{#1}} \newcommand{\given}{\mid} \newcommand{\green}[1]{\color{green}{#1}} \newcommand{\heads}{\outcome{heads}} \newcommand{\iid}{\overset{\text{\small{iid}}}{\sim}} \newcommand{\lik}{\mathcal{L}} \newcommand{\loglik}{\ell} \DeclareMathOperator*{\maximize}{maximize} \DeclareMathOperator*{\minimize}{minimize} \newcommand{\mle}{\textsf{ML}} \newcommand{\nimplies}{\;\not\!\!\!\!\implies} \newcommand{\orange}[1]{\color{orange}{#1}} \newcommand{\outcome}[1]{\textsf{#1}} \newcommand{\param}[1]{{\color{purple} #1}} \newcommand{\pgsamplespace}{\{\green{1},\green{2},\green{3},\purp{4},\purp{5},\purp{6}\}} \newcommand{\prob}[1]{P\left( #1 \right)} \newcommand{\purp}[1]{\color{purple}{#1}} \newcommand{\sign}{\text{Sign}} \newcommand{\spacecap}{\; \cap \;} \newcommand{\spacewedge}{\; \wedge \;} \newcommand{\tails}{\outcome{tails}} \newcommand{\Var}[1]{\text{Var}[#1]} \newcommand{\bigVar}[1]{\text{Var}\mkern-4mu \left[ #1 \right]} \]

Prof. Jeff Introduction!

• Born and raised in NW DC → high school in Rockville, MD
• University of Maryland: Computer Science, Math, Economics (2008-2012)

Grad School

• Studied abroad in Beijing (Peking University/北大) → internship with Huawei in Hong Kong (HKUST)

• Stanford for MS in Computer Science (2012-2014)
• Research Economist at UC Berkeley (2014-2015)

• Columbia for PhD[+Postdoc] in Political Science (2015-2023)

Dissertation (Political Science + History)

“Our Word is Our Weapon”: Text-Analyzing Wars of Ideas from the French Revolution to the First Intifada

Why Is Georgetown Having Me Teach This?

• Quanty things, but then PhD major was Political Philosophy (concentration in International Relations)
• What most interested me: unraveling history; Easy to get lost in “present-day” details of e.g. debiasing algorithms and fairness in AI, but these questions go back literally thousands of years!
• Pol philosophers distinguish “ancients” and “moderns” based on a crucial shift in perspective: ancients sought perfection, while Rousseau (1762) “took men [sic] as they are, and laws as they could be”.

Figure 1: Years spent questing in various dungeons of academia

• But is separation of ethics from politics possible? (Bowles 2016) Should we accept “human nature” as immutable/eternal? My answer: yes AND no simultaneously…

Dialectics

My Biases

• Upbringing: religious Jewish, right-wing (Revisionist Zionist) Republican environment
• “Encouraged” to emigrate to Israel for IDF service, but after learning history I renounced citizenship etc., family no longer big fans of me (Traumatic and scary to talk about, tbh 🙈)
• 2015-present: Teach CS + design thinking in refugee camps in West Bank and Gaza each summer (Code for Palestine)
• Metaethics: Learn about the world, challenge+update prior beliefs (Bayes’ rule!); I hope to challenge+update them throughout semester, with your help 🙂

On the One Hand…

On the Other Hand…

Remembering Why It Matters

Rules of Thumb

• Ask questions about power \(\leadsto\) inequities, but especially about structures/processes that give rise to them!
• “Philosophers have hitherto only tried to understand the world; the point, however, is to change it.” (Marx 1845)
• Dialectical implication: the more we understand it the better we’ll be at changing it

Ricardo Morales Art Studio / My office

Ethics as an Axiomatic System

Axiomatics

• Popular understanding of math: Deals with Facts, statements are true or false
  • Ex: \(1 + 1 = 2\) is “true”
• Reality: No statements in math are absolutely true! Only conditional statements are possible to prove!
• We cannot prove atomic statements \(q\), only implicational statements: \(p \implies q\) for some axiom(s) \(p\).
  • \(1 + 1 = 2\) is indeterminate without definitions of \(1\), \(+\), \(=\), and \(2\)!
  • (Easy counterexample for math/CS majors: \(1 + 1 = 0\) in \(\mathbb{Z}_2\))

Steingart (2023)

Example: \(1 + 1 = 2\)

• How it’s taught: this is a rule, and if you don’t follow it you will be banished to eternal hellfire
• How it’s proved: \(ZFC \implies [1 + 1 = 2]\), where \(ZFC\) stands for the Zermelo-Fraenkel Axioms with the Axiom of Choice!

Whitehead and Russell (1910), p. 83. See here for page in context

Proving \(1 + 1 = 2\)

(A non-formal proof that still captures the gist:)

• Axiom 1: There is a type of thing that can hold other things, which we’ll call a set. We’ll represent it like: \(\{ \langle \text{\textit{stuff in the set}} \rangle \}\).
• Axiom 2: Start with the set with nothing in it, \(\{\}\), and call it “\(0\)”.
• Axiom 3: If we put this set \(0\) inside of an empty set, we get a new set \(\{0\} = \{\{\}\}\), which we’ll call “\(1\)”.
• Axiom 4: If we put this set \(1\) inside of another set, we get another new set \(\{1\} = \{\{\{\}\}\}\), which we’ll call “\(2\)”.
• Axiom 5: This operation (creating a “next number” by placing a given number inside an empty set) we’ll call succession: \(S(x) = \{x\}\)
• Axiom 6: We’ll define addition, \(a + b\), as applying this succession operation \(S\) to \(a\), \(b\) times. Thus \(a + b = \underbrace{S(S(\cdots (S(}_{b\text{ times}}a))\cdots ))\)
• Result: (Axioms 1-6) \(\implies 1 + 1 = S(1) = S(\{\{\}\}) = \{\{\{\}\}\} = 2. \; \blacksquare\)

How Is This Relevant to Ethics?

(Thank you for bearing with me on that 😅)

• Just as mathematicians slowly came to the realization that

\[ \textbf{mathematical results} \neq \textbf{(non-implicational) truths} \]

• I hope to help you see how

\[ \textbf{ethical conclusions} \neq \textbf{(non-implicational) truths} \]

• When someone says \(1 + 1 = 2\), you are allowed to question them, and ask, “On what basis? Please explain…”.
  • Here the only valid answer is a collection of axioms which entail \(1 + 1 = 2\)
• When someone says Israel has the right to defend itself, you are allowed to question them, and ask, “On what basis? Please explain…”
  • Here the only valid answer is an ethical framework which entails that Israel has the right to defend itself.

Axiomatic Systems: Statements Can Be True And False

• Let \(T\) be the sum of the interior angles of a triangle. We’re taught \(T = 180^\circ\) is a “rule”
• Euclid’s Fifth Postulate \(P_5\): Given a line and a point not on it, exactly one line parallel to the given line can be drawn through the point.

\(P_5 \implies T = 180^\circ\) (Euclidean Geometry)

\(\neg P_5 \implies T \neq 180^\circ\) (Non-Euclidean Geometry)

Ethical Systems: Promise-Keeping

• Scenario: You just baked a pie, and you promised your friend you’d give them the pie. You’re walking over to the friend’s house to give them the pie.
• Suddenly, you turn the corner to encounter a hostage situation: the hostage-taker is going to kill their hostage unless someone gives them a pie in the next 30 seconds
• Do you give the hostage-taker the pie?

Consequentialist Ethics \(\implies\) Yes

• To be ethical is to weigh consequences of your actions
• The positive consequences of giving the pie to the hostage-taker (saving a life) outweigh the negative consequences (breaking your promise to your friend)
• (Ex: Utilitarianism, associated with British philosopher Jeremy Bentham)

Deontological Ethics \(\implies\) No

• To be ethical is to live by rules which you would want everyone to follow.
• As a rule (a “categorical imperative”), you must not break promises. (Breaking this rule \(\implies\) others can also “pick and choose” when to honor promises to you)
• (Ex: Kantian Ethics, associated with German philosopher Immanuel Kant)

Making and Evaluating Ethical Arguments

Descriptive vs. Normative

bin Laden (2005)

Descriptive Statement: “Bin Laden attacked us because we had been bombing Iraq for 10 years”	Normative Statement: “Bin Laden attacked us because we had been bombing Iraq for 10 years, and that is a good justification”
Descriptively True (empirically verifiable)	Normatively True (entailed by axioms + descriptive facts) in some ethical systems, Normatively False (not entailed by axioms + descriptive facts) in others

The Is-Ought Distinction

Hume on Is vs. Ought (Hume 1739)

• the author proceeds for some time in the ordinary way of reasoning
• suddenly, instead of the usual copulations of propositions is and is not,
• I meet with no proposition that is not connected with an ought, or an ought not.
• This change is imperceptible; but is, however, of the last consequence.

Descriptive (Is)	Normative (Ought)
Grass is green (true)	Grass ought to be green (?)
Grass is blue (false)	Grass ought to be blue (?)

What Happens When We Confuse The Two?

• Makes it impossible to “cross the boundary” between your own and others’ beliefs
• Collective welfare: Bad on its own terms (see: wars, racism, etc.)
• Self-interest: Prevents us from convincing other people of our arguments

Geertz (1973)

Collective vs. Self-Interest

• Good for collection of people \(\; \nimplies\) good for each individual person! (😰)
• \(p\) = Unions improve everyone’s workplace conditions, whether or not they pay dues
• \(q\) = Union dues are voluntary
• \(p \wedge q \implies\) I can obtain benefits of unions without paying
• \(\implies\) Individually rational to not pay dues
• (Think also about how this applies to climate change policy) 🤔

Olson (1965)

Modeling Individual vs. Societal Outcomes

• Individual Perspective: Individual \(i\) chooses whether or not to pay union dues

\(\implies\) Social Outcome: No Union

\(\implies\) Social Outcome: Union Possible

Key reading: Schelling (1978), Micromotives and Macrobehavior

Takeaway for Policy Whitepapers

• You cannot (just) say, “doing \(x\) will be better for society”
• You must also justify benefits to individuals, or at minimum, the individual organization and its stakeholders!
• (Is this a normative or descriptive claim?)

Ethical Issues in Data Science

• Data Science for Who?
• Methodological Individualism
• Operationalization
• Fair Comparisons
• Implementation

Data Science for Who?

• What are the processes by which data is (or is not) measured, recorded, and distributed?
• Who are the agents doing or not-doing these things?

The Library of Missing Datasets. From D’Ignazio and Klein (2020)

Example: Measuring “Freedom” and “Human Rights”

Methodological Individualism

• Atoms exhibit properties which are fruitful for understanding the physical world: we can study these properties as “building blocks” \(\leadsto\) interactions among atoms with various properties give rise to higher-level physical “things” (molecules, chemicals, cells, organisms)
• Individuals exhibit properties which are fruitful for understanding the social world: we can study these properties as “building blocks” \(\leadsto\) interactions among individuals with various properties give rise to higher-level social processes (dyads, groups, institutions)

For overthinkers: quarks \(\leadsto\) atoms as mental modules \(\leadsto\) individuals 😉 (Fodor 1983)

Structural Domination: The Grapes of Wrath

But… I built it with my hands! Straightened old nails to put the sheathing on!

It’s not me. There’s nothing I can do. I’ll lose my job if I don’t do it. And look—suppose you kill me? They’ll hang you, and long before you’re hung there’ll be another guy here, he’ll bump the house down. You’re not killing the right guy.

That’s so… Who gave you orders? I’ll go after him. He’s the one to kill.

You’re wrong. He got his orders from the bank. ‘Clear those people out or it’s your job.’

Well, there’s a president of the bank. A board of directors. I’ll fill up my rifle, head to the bank.

The bank gets orders from the East. ‘Make the land show profit or we’ll close you up.’ We’re sorry. It’s not us. It’s the monster. The bank isn’t like a man.

Yes, but the bank is only made of men!

No, you’re wrong there—quite wrong. The bank is something else than men. It happens nowadays that every man in a bank hates what the bank does, and yet the bank does it. The bank is something more than men, I tell you.

I got to figure… We all got to figure. There’s some way to stop this. There’s got to be some way to stop this. It’s not like lightning or earthquakes. We’ve got a bad thing made by men, and by God, isn’t that something we should be able to change? (Steinbeck 1939)

Ontology: Individuals and Structures

Methodological Individualism and Structural Domination!

No longer much preoccupied with such crudities as ‘conspiracy theory’, [progressives] have become quite monolithic in attributing all things negative to handy abstractions like ‘capitalism’, ‘the state’, ‘structural oppression’, and ‘hierarchy’. Hence they have been able to conjure what might be termed the ‘miracle of immaculate genocide’, a form of genocide, that is, in which there are no actual perpetrators and no one who might ‘really’ be deemed culpable […] The parallels between this ‘cutting edge’ conception and the defense mounted by postwar Germans are as eerie as they are obvious. (Churchill 2003)

Giddens (1979)

Operationalization

• Think of claims commonly made based on “data”:
  • Markets create economic prosperity
  • A glass of wine in the evening prevents cancer
  • Policing makes communities safer
• How exactly are “prosperity”, “preventing cancer”, “policing”, “community safety” being measured? Who is measuring? Mechanisms for feedback \(\leadsto\) updates?

Stiglitz, Sen, and Fitoussi (2010)

What Is Being Compared?

Apples	Oranges	Pears
Polities w/250-500M people (US ~335M, UP ~250M, EU ~450M)	Polities w/11M people in the Caribbean (Cuba, Haiti, Dominican Republic)	Polities w/over 1 billion people (China ~1.4B, India ~1.4B, Africa ~1.4B, ⬆️+⬇️ America ~1B)
Democracies (US)	Democracies til they democratically elected someone US didn’t like (Iran, Guatemala, Chile)	Non-democracies which brutally repress democratic movements w/US arms (Saudi Arabia)
Colonizing polities (US)	Polities colonized by them (Philippines)	Non-colonized polities (Ethiopia, Thailand)
Polities w/infrastructure built up over 250+ yrs via slave labor (US 🇺🇸)	Polities populated by former slaves (Liberia 🇱🇷)	Polities that paid reparations to descendants of [certain] enslaved groups (Germany)
Polities independent since 1776 (US)	Polities independent since 1990 (Namibia)	Non-self-governing polities (Puerto Rico, Palestine, New Caledonia)
Polities enforcing a 60 yr embargo on Cuba (US)	Polities with a 60 yr embargo imposed on them by US (Cuba)	Polities without a 60 yr embargo imposed on them by US (…)

How Are They Being Compared?

• What metric? Over what timespan?
• What unit of obs \(\leadsto\) agg function \(\leadsto\) level of aggregation?

Drèze and Sen (1991)

…There is Still Hope! I Promise!

• Fair Comparison through Statistical Matching:
• Lyall (2020): “Treating certain ethnic groups as second-class citizens […] leads victimized soldiers to subvert military authorities once war begins. The higher an army’s inequality, the greater its rates of desertion, side-switching, and casualties”

Matching constructs pairs of belligerents that are similar across a wide range of traits thought to dictate battlefield performance but that vary in levels of prewar inequality. The more similar the belligerents, the better our estimate of inequality’s effects, as all other traits are shared and thus cannot explain observed differences in performance, helping assess how battlefield performance would have improved (declined) if the belligerent had a lower (higher) level of prewar inequality.

Since [non-matched] cases are dropped […] selected cases are more representative of average belligerents/wars than outliers with few or no matches, [providing] surer ground for testing generalizability of the book’s claims than focusing solely on canonical but unrepresentative usual suspects (Germany, the United States, Israel)

Does Inequality Cause Poor Military Performance?

Covariates	Sultanate of Morocco Spanish-Moroccan War, 1859-60	Khanate of Kokand War with Russia, 1864-65
\(X\): Military Inequality	Low (0.01)	Extreme (0.70)
\(\mathbf{Z}\): Matched Covariates:
Initial relative power	66%	66%
Total fielded force	55,000	50,000
Regime type	Absolutist Monarchy (−6)	Absolute Monarchy (−7)
Distance from capital	208km	265km
Standing army	Yes	Yes
Composite military	Yes	Yes
Initiator	No	No
Joiner	No	No
Democratic opponent	No	No
Great Power	No	No
Civil war	No	No
Combined arms	Yes	Yes
Doctrine	Offensive	Offensive
Superior weapons	No	No
Fortifications	Yes	Yes
Foreign advisors	Yes	Yes
Terrain	Semiarid coastal plain	Semiarid grassland plain
Topography	Rugged	Rugged
War duration	126 days	378 days
Recent war history w/opp	Yes	Yes
Facing colonizer	Yes	Yes
Identity dimension	Sunni Islam/Christian	Sunni Islam/Christian
New leader	Yes	Yes
Population	8–8.5 million	5–6 million
Ethnoling fractionalization (ELF)	High	High
Civ-mil relations	Ruler as commander	Ruler as commander
\(Y\): Battlefield Performance:
Loss-exchange ratio	0.43	0.02
Mass desertion	No	Yes
Mass defection	No	No
Fratricidal violence	No	Yes

Bro Snapped

(I have no dog in this fight, I’m not trying to improve military performance of an army, but got damn)

Implementation

From D’Ignazio and Klein (2020), Ch. 6 (see also)

From Lerman and Weaver (2014) (see also)

Fairness… 🧐

Figure 2: From Lily Hu, Direct Effects: How Should We Measure Racial Discrimination?, Phenomenal World, 25 September 2020

Figure 3: From Kasy and Abebe (2021)

…And INVERSE Fairness 🤯

From Machine Learning What Policymakers Value (Björkegren, Blumenstock, and Knight 2022)

Ethical Issues in Applying Data Science

Facial Recognition Algorithms

Facia.ai (2023)

Wellcome Collection (1890)

Ouz (2023)

Wang and Kosinski (2018)

Large Language Models

When to retain biases…

…and when to debias

Figure 4: From Schiebinger et al. (2020)

Figure 5: From DeepLearning.AI’s Deep Learning course

Military and Police Applications of AI

Ayyub (2019)

McNeil (2022)

Your Job: Policy Whitepaper

• So… is technology/data science/machine learning “bad” in and of itself, or a tool to be wielded for both “good” and “bad” uses?
• How can we curtail uses of some kinds and/or encourage other uses?
• If only we had some sort of… institution… for governing its use in society… some sort of… govern… ment?

From Week 7 Onwards, You Work At A Think Tank

Morozov (2015)

From Ames (2014)

References

Ames, Mark. 2014. “The Techtopus: How Silicon Valley’s Most Celebrated CEOs Conspired to Drive down 100,000 Tech Engineers’ Wages,” January. http://web.archive.org/web/20200920042121/https://pando.com/2014/01/23/the-techtopus-how-silicon-valleys-most-celebrated-ceos-conspired-to-drive-down-100000-tech-engineers-wages/.

Ayyub, Rami. 2019. “App Aims to Help Palestinian Drivers Find Their Way Around Checkpoints.” The Times of Israel, August. https://www.timesofisrael.com/app-aims-to-help-palestinian-drivers-find-their-way-around-checkpoints/.

bin Laden, Osama. 2005. Messages to the World: The Statements of Osama Bin Laden. Verso Books.

Björkegren, Daniel, Joshua E. Blumenstock, and Samsun Knight. 2022. “(Machine) Learning What Policies Value.” arXiv. https://doi.org/10.48550/arXiv.2206.00727.

Bowles, Samuel. 2016. The Moral Economy: Why Good Incentives Are No Substitute for Good Citizens. Yale University Press.

Churchill, Ward. 2003. On the Justice of Roosting Chickens: Reflections on the Consequences of U.S. Imperial Arrogance and Criminality. AK Press.

D’Ignazio, Catherine, and Lauren F. Klein. 2020. Data Feminism. MIT Press.

Drèze, Jean, and Amartya Sen. 1991. “China and India.” In Hunger and Public Action, 0. Oxford University Press. https://doi.org/10.1093/0198283652.003.0011.

Facia.ai. 2023. “Facial Recognition Helps Vendors in Healthcare.” Facia.ai. https://facia.ai/blog/facial-recognition-healthcare/.

Fodor, Jerry A. 1983. The Modularity of Mind. MIT Press.

Geertz, Clifford. 1973. The Interpretation Of Cultures. Basic Books.

Giddens, Anthony. 1979. Central Problems in Social Theory: Action, Structure, and Contradiction in Social Analysis. University of California Press.

Hume, David. 1739. A Treatise of Human Nature: Being an Attempt to Introduce the Experimental Method of Reasoning Into Moral Subjects; and Dialogues Concerning Natural Religion. Longmans, Green.

Kasy, Maximilian, and Rediet Abebe. 2021. “Fairness, Equality, and Power in Algorithmic Decision-Making.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 576–86. FAccT ’21. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3442188.3445919.

Lerman, Amy E., and Vesla M. Weaver. 2014. Arresting Citizenship: The Democratic Consequences of American Crime Control. University of Chicago Press.

Lyall, Jason. 2020. Divided Armies: Inequality and Battlefield Performance in Modern War. Princeton University Press.

Marx, Karl. 1845. Thesen über Feuerbach. Stuttgart: J. H. W. Dietz. https://de.wikisource.org/wiki/Thesen_%C3%BCber_Feuerbach.

McNeil, Sam. 2022. “Israel Deploys Remote-Controlled Robotic Guns in West Bank.” AP News, November. https://apnews.com/article/technology-business-israel-robotics-west-bank-cfc889a120cbf59356f5044eb43d5b88.

Morozov, Evgeny. 2015. “Socialize the Data Centres!” New Left Review, no. 91 (February): 45–66.

Olson, Mancur. 1965. The Logic of Collective Action. Harvard University Press.

Ouz. 2023. “Google Pixel 8 Face Unlock Vulnerability Discovered, Allowing Others to Unlock Devices.” Gizmochina. https://www.gizmochina.com/2023/10/16/google-pixel-8-face-unlock/.

Rousseau, Jean-Jacques. 1762. The Social Contract. Geneva: J. M. Dent.

Schelling, Thomas C. 1978. Micromotives and Macrobehavior. Norton.

Schiebinger, Londa, Ineke Klinga, Hee Young Paik, Inés Sánchez de Madariaga, Martina Schraudner, and Marcia Stefanick. 2020. “Machine Translation: Gendered Innovations.” http://genderedinnovations.stanford.edu/case-studies/nlp.html#tabs-2.

Steinbeck, John. 1939. The Grapes of Wrath. Penguin.

Steingart, Alma. 2023. Axiomatics: Mathematical Thought and High Modernism. University of Chicago Press.

Stiglitz, Joseph E., Amartya Sen, and Jean-Paul Fitoussi. 2010. Mismeasuring Our Lives: Why GDP Doesn’t Add Up. The New Press.

Wang, Yilun, and Michal Kosinski. 2018. “Deep Neural Networks Are More Accurate Than Humans at Detecting Sexual Orientation from Facial Images.” Journal of Personality and Social Psychology 114 (2): 246–57. https://doi.org/10.1037/pspa0000098.

Wellcome Collection. 1890. “Composite Photographs: "The Jewish Type".” https://wellcomecollection.org/works/ngq29vyw.

Whitehead, Alfred North, and Bertrand Russell. 1910. Principia Mathematica. Cambridge University Press.