<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Multiplicity — Papers</title>
    <link>https://multiplicity.dev/</link>
    <description>Long-form writing on behavior analysis and AI by David M. Cole.</description>
    <language>en-us</language>
    <atom:link href="https://multiplicity.dev/rss.xml" rel="self" type="application/rss+xml"/>
    <lastBuildDate>Thu, 23 Apr 2026 12:00:00 GMT</lastBuildDate>
    <item>
      <title>A machine that says sorry</title>
      <link>https://multiplicity.dev/papers/a-machine-that-says-sorry/</link>
      <guid isPermaLink="true">https://multiplicity.dev/papers/a-machine-that-says-sorry/</guid>
      <description><![CDATA[Anthropomorphic AI gives us familiar but flawed controls. The current risk is less that users believe the machine is sentient and more that the interface cues the wrong repertoire.]]></description>
      <pubDate>Thu, 23 Apr 2026 12:00:00 GMT</pubDate>
    </item>
    <item>
      <title>The field that taught machines to learn</title>
      <link>https://multiplicity.dev/papers/the-field-that-taught-machines-to-learn/</link>
      <guid isPermaLink="true">https://multiplicity.dev/papers/the-field-that-taught-machines-to-learn/</guid>
      <description><![CDATA[Behavior analysis, artificial intelligence, and the cost of not building. Reinforcement learning traces back to Thorndike and Skinner. The people building modern AI know this. Behavior analysis has not made that lineage central to its own account of AI — and outsiders are building the measurement layer around autism care without it.]]></description>
      <pubDate>Fri, 17 Apr 2026 12:00:00 GMT</pubDate>
    </item>
  </channel>
</rss>
