Hi! I'm Caleb Strait

I'm a cognitive scientist & data scientist specializing in human decision-making processes and their shortfalls. My doctoral work focused on applying machine learning techniques to neural data collected mid-decision. Professionally I build models to support medical decisions in healthcare and clinical trial settings. In my spare time I work on my conversation framework mobile app, Discourser, which I've designed from the ground-up to encourage opinion convergence among debate participants.

Discourser

A mobile app framework for productive discussions

Discourser uses a tree framework for the visualization of an argument’s structure. The argument begins, in this case, as a central statement to be accepted or rejected. The central statement can be given any number of sub-statements, which are each explicitly designated as a supporting statement or a dissenting statement. Each sub-statement can also have any number of its own supporting or dissenting sub-statements. This recursive structure allows participants to fully document the intricacies of not only the supporting and dissenting sub-statements pertaining to a central statement, but also to suss out the strength of each of those sub-statements through the same system as with the central statement: a series of supporting and dissenting sub-statements.
Discourser maintains agreement scores for each statement that is added to a conversation, allowing for full documentation of participants’ relevant opinions on each part of the argument. Similar to how online comment sections allow for the vote-based finding of the best text-based argument, agreement scores give us a way to sort out how convincing parts of the argument are separately from one another, as segmented via the tree framework.

Blog


3/27/2024: Reversing Votes with an Argument App

    “Did you hear about that law we’re voting on?” I didn’t have to strain to hear my wife, Jessica, in the next room. But her question, to our host and one of the friends we disagreed with politically, made me squirm. I agonized over whether I could stand to remain in earshot for much of this sudden conversational escalation, from pleasantries to something (anything) political. I love her for her social bravado, and yes, this was the plan, but I had to escape after 20 minutes of uncomfortable listening to the two of them speak of the generally-not-spoken-of.
    They argued for hours.
    This was pre-pandemic- sometime in the Trump presidency, and a time when I, like many Americans, still believed in the power of conversation to change someone’s mind on political topics of debate. Not that I had often shaken off my social anxiety enough to try in earnest. But this show of determination by Jessica eclipsed any previous attempts by either of us at wading into the political quagmire with someone whose voting preferences we hoped to reverse.
    That goal was new itself for the two of us: converting votes. It was galvanized in us by our educations and the mutual understanding that the status-quo rule, don’t talk about politics, leads to legislational gridlock. For me, reversing votes was a goal that I had enshrined in my mind and then hidden from at every chance I got. Watching my wife actually try and (as far as we know) fail to convince our friend to vote with us crushed my hope that it could even be done in such a politically divided country.
    She did everything right: mentioning sources, remaining friendly, highlighting areas of agreement, and clearly outlining the actual issues at stake in our friend’s impending vote. And our friend responded with popular talking points that Jessica argued against in turn. Phones were brought out, with a few headlines thrown about, but even in an hours-long conversation, who has time to read articles and do due diligence on sources? And who is actually willing to change such a fundamental belief over the course of a few hours? Should Jessica hope to monopolize our visits to this friend with political argument for the foreseeable future? How could we possibly reverse a vote given this dynamic?

    So we came up with a new goal: put code on mine and a friend’s iPhones that would allow us to track a set of sources relevant to a disagreement long enough to read them, come back with informed opinions and repeat the process through a series of related topics. This program needed to be on our phones, because arguing this way takes good advantage of idle time, which we spend a good degree more of on our phones than other devices. Additionally, my friends and I were already responsive to mobile notifications but not desktop notifications or any other such format.
    The idea that I could manage to put an app on even a few iPhones only seemed available to me because professionally I work as a data scientist writing statistical scripts. I quickly learned that the idea that these two types of programming had much overlap was pure hubris, and the goal of producing and using the tool before the 2020 presidential election, which felt more important to win than any before, became 2024, which holds that same accolade.
    Now in 2024, with the help of some very generous and technical friends, I understand that learning app development involves wrapping one’s head around a huge techstack: the application’s user interface, the database, the server, & the cloud service host. Over the past 6 years I’ve slowly scraped together bare-bones understandings of these elements, spun up basic versions of them, and connected them to each other- a sprawling web of buttons, text, text input boxes, and server functions that together could bring my dream of a technologically structured argument to fruition.
    The app needed to both make it easy to list a set of sources, but also to track a robust conversation about each of those sources. This ended up implemented like branching roots of a tree: where a central disputed statement like, “Caleb should vote for Mia next month” would have an arbitrary number of supporting or dissenting child statements: either links to sources, or more logic like, “Mia is very experienced” to be supported or dissented in turn by more sources. This recursive organization of sources is necessary these days as arguments often hinge on demonstrating the relative reliability of two conflicting sources, a process which can have an arbitrarily large number of fact-checking layers.
    This tree of statements and sources would be helpful for organizing ideas, which is great for laying out how you disagree with someone. But what about when someone starts to change their mind, even just about one topic? The app needed to closely track each users’ opinion on each statement and source, and make that information clear to every participant in the debate. This was implemented with continuously updatable ‘agreement scores’ on every statement, spanning from “Caleb completely disagrees with this,” through “Caleb mildly agrees with this,” and ending at “Caleb completely agrees with this.” To accommodate agreement scores, sources were converted into a link mode for statements- so every item in the tree was a statement of logic like, “Mia would be tough on crime,” but some such statements would have links to articles attached and tend to be phrased more along the lines of, “This article describes how Mia tends to be tough on crime.”

    After 6 years of free-time development, the app was complete and on both mine and our friend’s iPhones using Apple’s phone software testing framework. I had everything I needed to give Jessica’s argument the high-tech redo I had been planning for over half a decade. Admittedly the matchup was now between myself, so that I could demonstrate how the app was intended to be used, and a different friend of ours: the nicest American I know, who happens to disagree with me politically. Most of my friends were aware of Discourser at this point. When you have the same hobby project for years, it tends to come up in conversation. I figured I could find no better dispositioned partner with whom to iron out the app’s details. And a vote is a vote. I texted this friend, invoking the name I had given the app. “Hey, would you be willing to try a political argument on Discourser? You could even just play devil’s advocate,” I wrote, doubting that such a concession would change the convincingness of a debate I was so desperate to have. My friend mildly agreed,
    “Sure.”
    Debates in this format felt so different from a face-to-face conversation, where two dueling, yet very humanly meandering trains of thought were confined to a few hours or less. Arguing over Discourser felt instead like playing chess where each player is given multiple days to make their move, or in this case find the perfect rejoinder statement. Since any statement can prompt a whole sub-argument, we felt motivated to get the phrasing just right. It could be a slow process, but topics were handled succinctly, often what would be a long, involved examination of a particular topic was successfully reduced to three or four sentence-long statements that got at the heart of a topic’s relevance to parent statements. Arguably the most helpful aspect of the app was simply how it saved our place, including the exact form of the disagreement. Entire parts of the argument that used to be often rehashed, were glossed over as settled material, including both mine and my friend's opinions on each portion of a matter, even when returning to the discussion after a few weeks.
    Our debate, or ‘conversation’ as I call them in Discourser to lend them a more cooperative sound, is ongoing as I write this. The logical end would be when we both have the same agreement score on the top-level ‘central’ statement, “[Friend’s name] should vote for [my candidate] this November.” I’m not holding my breath for our friend to join me at “completely agrees with this,” anytime soon, but November is still a few months away. In the meantime, we’ve found significant areas of agreement, which have made us feel less at odds overall.
    The goal, however, remains the reversal of votes, although I’m not all the way back at the drawing board. Perhaps I would have more success with someone who is more of a centrist, but I’ve found that only my friends who have particularly partisan beliefs, one way or another, are willing to engage in any form of political discourse in the first place. I’m still working on finding the exception. And when I do, I hope they have an iPhone.