Following 6-step process outlined by Ingrid Gnerlich, Publisher for the Sciences at Princeton University Press:

“If you have a nascent idea for a book, here is a 6-step plan for using your stolen moments to get started:”

EDIT: Just noticed that PUP has a separate format for Book Proposal Guidelines, but it’s pretty much the same thing as what’s here (Not that PUP is the only place I’m considering.)*

1. Aims and scope

“Write a paragraph (aim for 300 words max) that describes the book and what you want it to accomplish.”

Working Title: Judgment Calling: Classification and its Discontinuities by Scott H. Hawley

(On the title: “Classification and its Discontinuities” is a “clever” nod to Sigmund Freud’s famous book and encapsulates the theme of the book, but the publisher may per more of a “How…” subtitle to make it catchier, e.g. “How Classification Rules the World” or something similarly modest ;-). Other possible titles instead of “Judgment Calling” might be “Fault Lines” – although there was a big “Fault Lines” book on US history since 1974 that came out last year – or “Sorting it Out”)

“What are rights?” “Does this person belong to that group?” “What is the good?” “Is this thing edible?” Machine learning models are taking over human decision making on unprecedented scales. A vast majority of these decisions involve classifications: What sort of thing is this? Will you be assigned to the “high risk” or “low risk” category? Such classifications then set policies on how people are treated. This is a popular-level book on classification in its many forms, from antiquity to modern machine learning systems, with a goal of helping readers make sense of what is happening today. It takes an interdisciplinary look at classification in fields such as biology, philosophy, library science, sociology, music, and psychology, using humor, historical anecdotes and illuminating interactive (embedded or online) illustrations. Humans have been making classifications for as long as there has been intelligence. Psychologists believe that the impulse to categorize is as innate as our capacity for language, because we naturally group the world into concepts in order to understand and reason about it. How are human classifiers and machine classifiers similar, and how are they different? It turns out that machine classifiers can behave in very non-intuitive ways, and are susceptible to ‘hacking’ or tampering in ways that would never fool a human. How does this translate into how our government, health care, and school systems will behave when outsourced to the machines (of major tech companies) at massive scale? Of particular interest are the potential impact for machine-based systems applied to areas previously thought of as the purview of the Humanities: textual analysis and moral categories. This is more than just an educational text, it is a response to the call for greater literacy on how decision-making systems operate, as well as a wonderful journey into the realization that classification is at the heart of what it means to be human.

Aside: What category does this book go in? “Yes.” Meaning, its categories are: Philosophy, Science, Computer Science, Humanities, Social Sciences, History, Artificial Intelligence, Media & Society, Psychology. The whole question of “where in the bookstore does this go” is begging the theme of the book. But let’s say “science.” Ish. I’m a physicist, so probably easiest to do “science.”

2. Chapter Outline

“Sketch out a 1-page outline of chapters, and write a sentence or two about each chapter, or a list of bullet points (whichever is easier for you), describing what that chapter will be about. Do you have an Introduction and Conclusion? If not, you should probably add them.”

FORWARD

by (ideally) Kate Crawford, if I can interest her (Or Cathy O’Neil, or…)

PART I: PRELIMINARIES

(Is “preliminaries” really the right word? It keeps up the “p” alliteration, but does it falsely imply that this content is somehow prior to my “main work,” or something that people might want to just skip over?)

CHAPTER 1: Bay Area Man, or “About this Book”

This chapter is a justification for the urgency, immediacy, relevance for ordinary people, and importance of this book. Heading-quote for this chapter is “Bay-area man proposes inventing a field of study that already exists” (Mark Riedl on Twitter, 2019).

  • Anecdote about how the power of ‘disruptive’ tech culture makes tech people think they can tread into/over others’ domains.
  • “Space Invaders” : On the successes of Bay Area Men, beating human experts with machine learning models, in a variety of domains.
  • Problems from the Bay Area Man mindset: injustice, content moderation issues, democracy?
    • In particular, the impact of classification, being on one side of a threshold or not
  • The inderdisciplinarity of this book, and the importance for historical context: what’s new vs. what’s old re. classification by humans vs. by machines (and Machine Learning in particular).

CHAPTER 2: SAMPLE CHAPTER: One Of These Things Is Not Like The Others

Autobiographical. Justification for my interest, and for me as the person to write this book.

  • Story of my time at Oxford, i.e., how I – a physicist – got interested in the topic of classification.

  • Set-up for next chapter: Which field “Owns” Classification? (And what business does a physicist have in writing on this?) Is it the librarians / information theorists (i.e. the Society for Knowledge Organization), the philosophers, biologists,…? The ‘overarching’ view of this book is sure to irritate specialists in these fields, but since the machines are being applied in a cross-disciplinary way, we as humans need a cross-disciplinary look. Let’s explore these fields in the next chapter, because unlike Bay Area Man, we want to understand history/context.

CHAPTER 3: A History of Sorts

This gives a historical overview of people’s thoughts about classification, from fields such as philosophy, biology, library science, medicine, sociology,…will save much of the psychology details for the “How it Works” chapter, later. The goal will be to keep this interesting to the reader and not too long; telling the reader some interesting historical points and important ideas but NOT trying to cover everything that’s ever been said on classification.)

  • Interactive graphical timeline

  • Key figures: Aristotle, Kant, Nietzsche, Wittgenstein, Eleanor Rosch, Isaiah Berlin, MacIntyre, Parrochia, Mai, and more.

  • …a bit about classifying people……And bureaucracy, and automation….

PART II: PERSPECTIVES

CHAPTER 4: SAMPLE CHAPTER: Noticing a Pattern: Classification as Pattern Recognition

  • Intro
  • Interactive Demo: If the Shape Fits Pass It
  • Learning to See
  • You Spot It, You Got It
  • Interactive Demo: Signal Correlations
  • What Do Computer Vision Classifiers ‘See’?
  • Broader Observations

CHAPTER 5: What Will We Do With This? Classification as Policy

‘Technically’ for my treatment, classification precedes policy yet feeds directly into it; policies are based on classifications. (Academic question: is policy “supervenient” on Classification?). Not all policies are based on classifications, but many (most?) are.

  • Example from Cybersecurity: Risk score (1 through 5) sets what you do in response. More famous example: DoD’s DEFCON scale.

  • Medicine: Diagnosing (classifying) a disease determines treatment

  • Law: Guilty / Not guilty is a classification, guilty-of-what determines available sentencing options. Classifying evidence as inadmissible means it won’t influence proceedings.

  • Teaching: Getting an F means you have to repeat the course.

  • Labeling you a Nazi means I’m justified in punching you. ;-)

  • Labeling abortion as murder means….?

  • Labeling my internet access as a “right” means you have to pay for it? ;-)

    TODO: this reminds me that I don’t really have a spot for my writings on the contentious nature of labels, i.e. “Labels Stick”, etc. Hmmm.

CHAPTER 6: Classification as Intervention

This is where the “Classification and its Discontinuties” subtitle really gets explored, re. the consequences of classification, i.e. being on one side vs. the other right near a decision boundary.

  • Effects of Criminal Risk Assessment - people classified as “high risk” end up with much bigger problems than were “just short” of getting dubbed “high risk.” Interview (criminal risk assessment data scientist & Nashville friend/colleage) Vienna Thompkins.
  • Regression Discontinuity Design - used to show effects of, e.g. merit scholarship programs.

CHAPTER 7: Classification as Power

  • …the whole Kate Crawford thing ;-)
  • TODO: fill this in. Publishers don’t know what ^that means and/or how I’d provide my own take on it.
  • Who gets to supply the labels? Borrow from my FaithTech Writing Contest essay.
  • Representation

PART III: PRACTICALITIES

CHAPTER 8: They Come In Two Classes: Binary Classification

  • Description of basic classifications in many fields

  • List of famous song lyrics involving binary classification tasks

  • Interactive Exercises

CHAPTER 9: Kinds of Kinds: Variations on Classification

  • Kinds: Grouping and Clustering

  • Orderings, Thresholds, Quotas

  • Boundaries, Territories

CHAPTER 10: How Classification Happens

We covered some of this already, but it’s important to go into greater detail.

  • By Humans: Psychology of classification

  • By Machines — big thrust of the book, re. my interest in educating the public, lots of interactive elements

CHAPTER 11: Our Bad: Misclassification

We’ve already talked about misclassification by machines as a motivating “problem” justifying the urgency of this book, and have touched on it some in humans in “Noticing a Pattern,” but a deeper dive into aspects of misclassification (by humans & machines) proves fascinating.

  • (start with self-driving car accidents)

  • Friend or Foe

  • Adversarial Attacks

PART IV: POSSIBILITIES

CHAPTER 12: Yea But What If: Counterfactuals

TODO: Does this really merit an entire chapter? Could maybe merge 10 & 11 into “Possible Futures” or “Possible Worlds”

  • Causal modeling

  • Counterfactuals as explanations of classifications

CHAPTER 13: Rise of the Machines?

  • Looking toward the future – which is the present.

  • Thoughts on future of work, gig economy,… Note that Google just (Sept 22 2020) brought BACK human moderators for YouTube

3. Readership and Market

Write a sentence to describe (realistically) the readership to which your book will naturally appeal. List a few other books on the market that are comparable in terms of subject matter and intended readership. Describe how your book is distinctive in comparison to each.

Readership

“Lay” intellectuals interested in technology & society, philosophy, history, computer science, machine learning and its applications, fairness and justice. People who read “high-popular” or “popular-academic” level science-y books, e.g. by Malcolm Gladwell, Steven Pinker, Brian Greene. My book would also be suitable for students.

Not for: Fully-academic publishing such as Routledge. Although not aimed specifically at scholars, the interdisciplinary nature of this book implies that scholars from various fields will find this a helpful way of expanding their “breadth.” But my book is more about exploring & educating, and forming interdisciplinary connections than about trying to challenge the status quo in a field or advance a new paradigm – so, e.g., not Epstein’s The Ant Trap (OUP Phil. of Sci. Series), or Hossenfelder’s Lost in Math (Basic Books, 2020).

Similar books

  • Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy by Cathy O’Neil (Crown Books – now Crown Publishing, 2016, 272 pages): Now-classic book, bestseller. Wonderfully, accessibly written. Largely deals with financial models, rankings, and feedback loops. Widely read, also used as a text in some low-level college classes. Difference to mine: More about “models” than classification per se, no historical overview, no interactivity, hardly anything on methods AFAIK, more about (unintended or unjust) effects that how things work. Crown’s Press Release for the book

  • Hello World: Being Human in the Age of Algorithms by Hannah Fry (W.W. Norton / Penguin, 2018, also 272 pages).

    Excellent, readable bestseller explains aspects of how algorithms work and their implications in many areas of society. It’s read popularly but also used as a text for low-level college courses (or at least my courses!). Difference to mine: Only historical/philosophical context is from 1950’s onward (a little bit of earlier stuff, not much), no interactivity, methods are described more vaguely compared to the “how it works” that my book will have.

  • The Creativity Code: Art and Innovation in the Age of AI by Macus du Sautoy (2019, Belknap Press: An Imprint of Harvard University Press). I wrote a review of this book at the request of the tech editor for PSCF, which they published. du Sautoy is a mathematician at Oxford & popularizer of science. Accessible and engaging author, combines historical anecdotes, personal/autobiographical bits, interviews with top people. If this wasn’t officially a “bestseller” it was pretty close! Like me, du Sautoy is combining some of his own takes but giving readers and educational tour of a (set of) field(s) that are not his wheelhouse. Like my book he does give historical context (e.g., Ada Lovelace, Mozart), but unlike my book his doesn’t offer any interactivity.

  • distill.pub is not a book but rather an online journal for review articles on topics in machine learning, featuring (beautiful!) embedded, interactive graphs – what they call “reactive diagrams.” Run by consortium of Google Brain, OpanAI & others; one of main editors is Chris Olah. This was a big inspiration in conceiving of this book, e.g. their interactive playground for t-SNE. Aimed toward educating clarifying things for researchers, not common folk. Not sure if they’d sponsor a book, but might offer encouragement. Might be worth borrowing some of their infrastructure, cf. how to prepare a distill article. Have thought about writing to Olah for any sort of tips, connections,…not sure yet.

…Those are the main similar books I have in mind. Writing about AI and/or “algorithms & society” has been a hot topic over the past couple years, with so many books it’s impossible to keep track, e.g.:

  • Algorithms of Oppression by Safiya Noble (NYU Press, 2018): haven’t read it. probably should ;-)
  • The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power by Shoshanna Zuboff (Public Affairs Books, 2019) Bestseller.
  • The Age of AI: Artificial Intelligence and the Future of Humanity by Jason Thacker (2020, Zondervan). Written for church-folk. My book is intended for a wide “secular” audience. (Thacker is in Nashville & is an acquaintance / colleague.)
  • Technically Wrong: Sexist Apps, Biased Algorithms, and Other Threats of Toxic Tech by Sara Wachter-Boettcher (2017) (Haven’t looked at it, but I follow her on Twitter.)

4. Revise

“Ask yourself if your chapters and the ideas within each of them are organized in such a way that they accomplish what your aims-and-scope paragraph states. Is there anything that would make your book even more distinctive in light of other books on the market? Be honest with yourself. Revise.”

In progress.

My interactive exercises are distinctive.

Revisions such as:

  • Combined what was Chapter 3 (“Space Invaders”) into Chapter 1; it keeps continuity and provides justification for the claims earlier in chapter 1.
  • Added “Parts” structure on top of chapters.

5. Revise Again

“Now ask yourself, how long do you think your book will be, approximately? Can you pare it down and still accomplish what your aims-and-scope paragraph describes? If so, do that. Long books are wonderful, but good things can also come in little packages.”

Shooting for “272” pages. ;-) Because my two favorite books, O’Neil’s & Fry’s were exactly that long. And ~270 seems to be the length of many popular books in my market. Note that my book would contain illustrations & such which take up space. So rather than 70,000 words, I’d likely be closer to 50,000 words.

I think that length is good. Shorter could work, we’ll see as I revise.

6. Timeline

“At last, you’ll come to the million-dollar question: how long do you think would it take to write the book you’ve outlined?”

4-6 months? Some days I do 1000 words/day, sometimes it takes 3 days to write that much. Interactive elements take 2-3 days to write, 2-3 more to revise.

7. More / Other

Beyond Gnerlich’s instructions, I’ve seen other advice that I should also plan out…

Publishers to approach

  • The popular arm of a university press might be suitable, e.g. Princeton, Oxford, NYU. These could be approached without an agent.
  • If I had an agent, then they would know. Maybe W.W. Norton, Crown, Bloomsbury (Merchants of Doubt, 2010)… Note: Seems that Basic Books (e.g. Hossenfelder’s book, 2020) is big on “shap[ing ]public debate,” which is not what I’m trying to do. Love their authors though: Turkle, Hofstadter, Pinker? Get out of town! And yet Basic just published Kocharski’s Rules of Contagion which is pure exposition/exploration – “an essential guide to modern life” and I am totally stealing that. ;-)

Marketing Plan

(Confused re. if this is necessary: A few authorship guides I read said that even for non-fiction we need this, but others say no that’s only for novelists. Seems to depend on publisher / agent.)

  • The book needs a website. e.g. to host interactive content to accompany the print edition. Once the title is formalized, I’ll reserve the domain name. Will change www.scotthawley.com so that it’s not just about my music. I’m on Twitter already & have ~900 followers. My professor page at Belmont will include a link to the book. Could even have a “FAQ” page on the book website.
  • Pre-release a chapter via my blog, to drum up anticipation for the book?
  • Sharing a sample chapter to appear in some periodical: Aeon? Quanta? Nautilus? The New Atlantis? The Atlantic? WIRED? Scientific American?
  • Speaking appearances on podcasts & radio, absolutely.
  • I could give lectures at universities, or science clubs, programming meetups, etc. “What about high schools?”…uh…not sure, maybe.
  • Maybe speak at book shops if they still do that? (Parnassus here in Nashville, but they’re more about ‘novelists’ than non-fiction, I think.)
  • Send out for book reviews to appear in book-conscious journals / periodicals such as The TLS, “The [whatever] Review of Books,” Books & Culture, PSCF (I reviewed Marcus du Sautoy’s The Creativity Code for them), The New Atlantis (The editor knows me, a bit),… Publisher probably does that already, but I could also ask.
  • Reach out to ML/AI people I know – Jeremy Howard, Joanna Bryson, Lilian Edwards, AIandFaith.org,… to help spread word about the book? Pop-physics authors I’m connected to: Briggs, Hossenfelder (Lost in Math, 2020, Basic Books)? Philosophers: McGrath. Maybe Olah? Similarly, notify Office of Communications at my university.
  • Keeping up other forms of authorship (FaithTech writing contest, chapter for Hector’s book, technical journal articles) to further demonstrate my author-worthiness? (useful for marketing?)