Partisan Motivated Reasoning Trumps Even Illusory Truth

class: center, middle, inverse, title-slide

.title[
# <span style="font-size: 90%;"> Partisan Motivated Reasoning Trumps Even Illusory Truth </span>
]
.author[
### <span style="font-size: 90%; line-height: 1.8"> Tiago Ventura (Georgetown University) </strong>, James Bisbee (Vanderbilt University), Sarah Graham (CSMaP-NYU) and Joshua A. Tucker (CSMaP-NYU)</span>
]
.date[
### <span style="font-size: 80%;"> <br> APSA 2024 </span>
]

---

layout: true

<div class="my-footer"><span>Tiago Ventura (Georgetown University) &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp  APSA 2024 </span></div>

---
class:middle
## Motivation

- Scholarly evidence that social media facilitates the spread of misinformation and polarizing content .midgrey[( Vosoughi et. al., 2018; Del Vicario et al., 2016; Aruguete, Calvo and Ventura, 2021; Eady et al., 2023)].

- Often see claims that conflate prevalence/virality features of misinformation on social media with real-world impact .midgrey[(Altay, Berriche and Acerbi, 2023; Wagner and Boczkowski, 2019)]

--
- **.red[But]**:
  - Exposure to misinformation is concentrated among small set of (heavily engaged) users .midgrey[(Grinberg et al., 2019, Budak et. al., 2024)]
  - Still a small part of informational environment for most users .midgrey[(Allen et. al., 2020, Guess et. al., 2021)]
  - Mixed evidence for causal effects of exposure beliefs and attitudes  .midgrey[(Eady et al., 2023, Allen et. al. 2024, Ventura et. al. 2024)]

---
class:middle

### Illusory Truth vs Directional Goals: Two distinct frameworks

-  **Pennycook, Cannon and Rand, 2018** on the effects of prior exposure to misinformation: 
  - "a single exposure increases subsequent perceptions of accuracy, both within the same session and after a week". 
  - This effect is not moderated by partisan motivations

---
class:middle

### **Our Contribution**:
 
**To which degree prior exposure to false news online trump partisan-motivated beliefs for political misinformation?**

- **Model:** Situate both frameworks in a common model of Bayesian belief formation AND 
- **Experiments:** Measure the effect of both frameworks on belief for political misinformation through a set of survey experiments.

---
class:middle
### Bayesian Model for Belief Formation

$$ 
`\begin{align*}
    \textbf{Prior:}~~\pi(\mu) \sim \mathcal{N}(\hat{\mu}_{i,0},\hat{\sigma}^2_{i,0})\\
    \textbf{Signal:}~~x \sim \mathcal{N}(\mu_{x},\hat{\sigma}^2_{i,x})
\end{align*}`
$$
$$
`\begin{align*}
    \pi_i(\mu|x) &\sim \mathcal{N}\left(\hat{\mu}_{i,0} + (\mu_x - \hat{\mu}_{i,0}) \left(\frac{\hat{\sigma}_{i,0}^2}{\hat{\sigma}_{i,0}^2 + \hat{\sigma}_{i,x}^2}\right),\frac{\hat{\sigma}_{i,0}^2\hat{\sigma}_{i,x}^2}{\hat{\sigma}_{i,0}^2 + \hat{\sigma}_{i,x}^2}\right)
\end{align*}`
$$

- critical parameters are `$\hat{\sigma}_{i,0}$` which measures the (inverse) credibility of the signal

- .red[PMR]: Source alignment <svg viewBox="0 0 512 512" style="height:1em;position:relative;display:inline-block;top:.1em;" xmlns="http://www.w3.org/2000/svg">  <path d="M504 256C504 119 393 8 256 8S8 119 8 256s111 248 248 248 248-111 248-248zm-448 0c0-110.5 89.5-200 200-200s200 89.5 200 200-89.5 200-200 200S56 366.5 56 256zm72 20v-40c0-6.6 5.4-12 12-12h116v-67c0-10.7 12.9-16 20.5-8.5l99 99c4.7 4.7 4.7 12.3 0 17l-99 99c-7.6 7.6-20.5 2.2-20.5-8.5v-67H140c-6.6 0-12-5.4-12-12z"></path></svg> credibility of the signal
  - .red[ITE]: Familiarity (or prior exposure) <svg viewBox="0 0 512 512" style="height:1em;position:relative;display:inline-block;top:.1em;" xmlns="http://www.w3.org/2000/svg">  <path d="M504 256C504 119 393 8 256 8S8 119 8 256s111 248 248 248 248-111 248-248zm-448 0c0-110.5 89.5-200 200-200s200 89.5 200 200-89.5 200-200 200S56 366.5 56 256zm72 20v-40c0-6.6 5.4-12 12-12h116v-67c0-10.7 12.9-16 20.5-8.5l99 99c4.7 4.7 4.7 12.3 0 17l-99 99c-7.6 7.6-20.5 2.2-20.5-8.5v-67H140c-6.6 0-12-5.4-12-12z"></path></svg> credibility of the signal

---
class:middle
## Design

- **Data:** Online survey fielded Qualtrics, with a nationally representative sample of Americans.

- **Design:** Modeled after previous work examining ”illusory truth effects” .midgrey[(Pennycook, Cannon and Rand, 2018; Lyons, 2023)]

- .red[Familiarization stage]: show .red[X] headlines with questions about participants' familiarization
  
  - .red[Distraction stage]: distract participants with survey questions
  
  - .red[Accuracy Stage]: show .red[X + Y] headlines with questions about accuracy beliefs
  
- **Measuring PMR**: In previous work, partisan leaning is manipulated only with content of the headlines .midgrey[(Pennycook, Cannon and Rand, 2018)]. We manipulate both the content and source (MSNBC, Democracy Now, Fox, Breitbart)

---
class:middle
## Experiments

---
class:middle
### Marginal Means ITE vs PMR for False Headlines

---
class:middle
### Marginal Effects ITE vs PMR for False Headlines

---
class:middle
### Effects of ITE vs PMR over Time for False Headlines

---
class:middle
### Effects of ITE vs PMR over Time for False Headlines with Warning Labels

---
class:middle
### Effects of ITE vs PMR over Time for TRUE Headlines

---
class:middle
## Discussion

-  Partisan motivated reasoning dominates illusory truth in the political realm

- PMR has effect order of magnitude larger than PMR
   - PMR moderates ITE with prior exposure being larger among concordant sources
   - ITE decays after a single day of exposure
   
   
- Notes of optimism: Warning labels do reduce both processes (ITE and PMR)

- Next steps:
  
  - New experiment adding non-partisan headlines as baseline
  - Proper measure for familiarity (separate prior knowledge to the effects of familiarization)
  - New experiment with fabricated (but partisan) headlines