Generative AI, Misinformation and 2024 Elections

.title[
# Generative AI, Misinformation and 2024 Elections
]
.author[
### Tiago Ventura <br> <br><span style="font-size: 80%;">McCourt School of Public Policy, Georgetown University </span>
]
.date[
### <span style="font-size: 80%;"> <br> GTPI Chat on AI & Misinformation <br> <br> 03/14/2024 </span>
]

---

<div class="my-footer"><span>Tiago Ventura &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp  GTPI Chat on AI & Misinformation</span></div>

---
## Open AI and ChatGPT

---
class:middle
### Google: AI and Elections

---
class: middle

#### .center[World Economic Forum’s 2024 Global Risks Report surveying 1500 experts and policy makers list “misinformation and disinformation” as the top global risk over the next two years]

---
class:middle
## To which extent AI will impact elections?

- Some warn of AI leading to a deluge of misinformation, especially on social media:
  - ["AI-generated content could soon account for 99% or more of all information on the internet"](https://www.axios.com/2023/07/10/ai-misinformation-response-measures)
  - ["Come near your and you will not realize"](https://www.wired.com/story/ai-generated-fake-news-is-coming-to-an-election-near-you/)
  - [Facilitates production of deep-fakes, micro-targetting and human-like misinformation.](https://www.wired.com/story/generative-ai-custom-disinformation/) 
  
- Others finds the fears understandable but largely overblown
   - Use recent social science research showing indicating 
      - [Misinformation exposure is overly concentrated](https://www.science.org/doi/full/10.1126/science.aau2706)
      - [Consumption is lead by highly partisan users](https://www.cambridge.org/core/journals/american-political-science-review/article/partisan-polarization-is-the-primary-psychological-motivation-behind-political-fake-news-sharing-on-twitter/3F7D2098CD87AE5501F7AD4A7FA83602)
      - [Effects of exposure do not translate mechanically into political outcomes](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4457400)

---
class:middle

## Talk today

The answer is probably in between these two camps. Idea for today:

- Bring some scientific evidence behind: 
   - **.red[Fears about AI are understandable but largely overblown]**
   - **.red[AI leading to a deluge of misinformation]**

- Discuss some open-questions on this topic and why we need more research

- Make sure you leave knowing this is a fairly complicated policy issue! 
   - And that simplification about the effects of AI might actually do most of the damage!

---
class:center, middle

![](https://media.giphy.com/media/v1.Y2lkPTc5MGI3NjExcTZoODY2aDd3czBlbTd0Ynlvd2g2cWg3cWJpbGRrZm04MnZ2c2cwOCZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw/xTiN0NZZsYXkGwbcJ2/giphy.gif)
---
class:center, middle, inverse

# Fears are understanble but largely overblown

---
class:middle

---
class: middle
## Three insights

---
class: middle
### .center[The average american consumes 7.5 hours of media per day, about 14% was related to news. Fake news comprises only 0.15% of Americans’ daily media diet]

[Evaluating the fake news problem at the scale of the information ecosystem, Allen et. al., 2020](https://www.science.org/doi/pdf/10.1126/sciadv.aay3539)
]

---
class: middle
### 89% of news URLs viewed on Facebook were from credible sources

[Cracking Open the News Feed: Exploring What U.S. Facebook Users Seeand Share with Large-Scale Platform Data, Guess et. al., 2021](https://journalqd.org/article/view/2586/2683)
]

---

---
class:middle 
### 1% of users represented 80% of exposure to Fake News on Twitter in 2016

.center[
<img src="output/gringer.png" width="80%" />
]
.center[[Grinberg et al., 2019](https://www.science.org/doi/pdf/10.1126/science.aau2706)]

---
### 90% of respondents didn’t share a single link from a fake news website during the election period.

[Less Than you Think, Guess et. al., 2019](https://www.science.org/doi/10.1126/sciadv.aau4586)

---
class: middle

## Three insights

---
class: middle

### Exposure to IRA Russian Misinformation on Twitter doesn’t lead to changes in attitudes

---
class: middle

#### Reducing exposure to misinformation on WhatsApp had null effects on beliefs for popular true and false stories weeks ahead of presidential in Brazil

]

[Beyond Traditional Feeds, Ventura et. al., 2023](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4457400)

---
class:center, middle, inverse

# How might AI change what we know about misinformation?

---
class:middle, center

## Production  <=>  Distribution of Misinformation

---
class: middle
#### AI Misinformation => Micro-targetting => Persuassion

---
class: middle

### The rise of TikTok, and the decline of social-graph based apps

---
class:middle
#### AI-generated misinformation could further erode trust: Liar's Dividend 
> Liar's Dividend: politicians or other public figures can—potentially credibly—claim that real news stories are merely fake news or deepfakes

.footnote[[Schiff et. al 2024](https://www.cambridge.org/core/journals/american-political-science-review/article/liars-dividend-can-politicians-claim-misinformation-to-evade-accountability/687FEE54DBD7ED0C96D72B26606AA073)]

---
class:middle
## Final Thoughts

Both camps are probably correct:

---
class: middle, inverse, center

# Thanks! And let's have a conversation!