CMU-HCII-20-102
Human-Computer Interaction Institute
School of Computer Science, Carnegie Mellon University



CMU-HCII-20-102

Adaptive Instruction:
Productivity of Civil Discourse Addressing Bias

Nicholas Diana

June 2020

Ph.D. Thesis

CMU-HCII-20-102.pdf
Pending


Keywords: NA


Civil discourse is our most basic form of civic engagement. In a democracy,it is our best tool for collectively answering a society's most fundamental question: "What shall we do?" While most of us have no doubt participated in political discussions, engaging in civil discourse that is productive (i.e., dialogue that "fosters democratic goals" [59]) can be substantially more difficult. For example, when both sides of an argument view their beliefs as part of their identity[43], debating the merits of those beliefs without calling into question the merits of the individuals who hold them can be challenging. Ideally, a careful discussant might find common ground, previously obscured by the trappings of tribalism, or, at the very least, foster a mutual under-standing of and respect for the values that inform beliefs they do not share. On the other hand, an unskilled discussant, perhaps one only interested inpersonal attacks or "winning arguments," will likely only further entrench each party in the views of their political tribe.

Engaging in productive civil discourse is a skill that needs to be explicitly taught, modeled, and practiced in the same way that students are taught and given opportunities to practice skills like finding the length of the hypotenuse or completing a Punnett square. Unfortunately, civic education often takes a backseat to so-called core subjects like math, science, and English. Learning about civics may be relegated to elective courses, and opportunities to practice civic reasoning skills only afforded to the members of the school's debate club. And even in these environments where we would expect civic education to be a central focus, the emphasis on argumentation and persuasion, while undoubtedly crucial to civi clearning, is insufficient and in some cases counter-productive to the type of dialogue thatengenders commonality and collective problem-solving.

In contrast to the status quo, we developed and tested the impact of a novel civic education intervention designed to provide students with 1) a better understanding of the values that shape their own beliefs and the beliefs of others, 2) opportunities to practice overcoming the biases that are born out our pre-existing beliefs, 3) an understanding of what makes civil discourse productive and examples of model civil discourse, and 4)opportunities to practice the skills that underpin productive civil discourse. We found that student performance on key civil discourse skills (e.g., value-identification, tribalism reduction) improved with practice. We also demonstrate, for the first time, that an AI-powered educational game that adapts instruction to a student’s specific values can be used to measure and, in some cases, reduce the impact of bias when reasoning about political beliefs. In addition to direct applications in civic technology, this new, value-adaptive instruction has implications for systems designed to mitigate bias and augment human reasoning

XXX pages

Thesis Committee:
Kenneth Koedinger (Co-Chair)
John Stamper (Co-Chair)
Jessica Hammer
Sharon Carver (Psychology)
Mathew Easterday (Northwestern University)

Jodi Forlizzi, Head, Human-Computer Interaction Institute
Martial Hebert, Dean, School of Computer Science



Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu