Anchoring in the past, tweeting from the present – Cognitive bias in journalists’ word choices

This study is really cool. It explored journalists’ language in their reporting of the US political campaign trail and what their word choices reveal about their cognitive mindsets – namely drawing on the system 1 / system 2 thinking concept and anchoring heuristic.

Journalists’ twitter posts were compared against newspaper articles and broadcasts (>220k articles, broadcasts and tweets were analysed).

I’ll give my thoughts about the relevance to safety practices in the discussion section.

System 1 thinking is a “rapid, low-effort, and intuitive way of thinking, often based on emotions and habits” whereas system 2 thinking involves processing info “more slowly, analytically, and systematically to arrive at a conclusion” (p1-2). People spend most of their time drawing on system 1, since it minimises cognitive effort and is time efficient for arriving at decisions.

Journalistic reporting of campaign trails is said to be an opportune area to evaluate system 1/2, since journalists are under time pressure to distil unfolding events and twitter allows fast dissemination of info. Journalists are expected to rely more heavily on system 1 thinking when posting on twitter, compared to more system 2 in formal media articles etc.

Journalism has been referred to as displaying “inherently “institutionalized heuristic decision making” (p3) because of system 1 thinking is embedded in journalistic practices due to the pressures and uncertainty around deadlines.

The authors cover research behind this topic:

Media users rely on system 1 thinking when navigating social media – drawing heavily on heuristics to process heavy streams of info
Based on the 2012 US political trail, the majority of political journalist twitter activity didn’t involve fact-checking
Similarly, clinician-patient consultation was equally affected. Under different time settings (pressures), when clinicians are under great time pressure, their interactions with patients involve changes in linguistic features such as more direct language, more emotion and providing less evidence for judgements.
Media, like many things, is influenced by perceptual biases/heuristics, like confirmation bias, recall bias or anchoring
One study found only half of the studied stories by journalists brought “alternative hypotheses” to the article, indicating confirmation bias. In other work, confirmation bias was evident in the article by the types of confirming sources used by the journalists (rather than using a balanced portfolio of sources) such that “journalists only referenced the sources that were critical of [their view] and omitted relevant information” (p4). Reporters selected sources and interpreted those sources in order to support their worldview.
Another study looked at the reporting of international news, based on Belgian media sources. It found that while natural disasters in Australia and US were portrayed as important and worthy of attention, which contrasted by language signifying the “distant ‘Others’” of Indonesia or Pakistan without cause for concern (p5).

These findings highlight the importance of sematic or grammatical features of text for understanding social systems and power hierarchies.

Results

Evidence was found of system 1 thinking journalists’ tweets, compared to news articles.
This also supported that journalists were less likely to draw on system 2 thinking in tweets (largely due to Twitter’s “emphasis on speed, spontaneity, and informal cultural milieu” (p13). Journalists used less analytical thinking terms, fewer quantifying/number terms with a strong difference between linguistic patterns.
Journalists used more positive and negative emotions in their tweets compared to their crafted news stories for papers; tweets were crafted more to evoke emotion.
Journalists showed more certainty in their opinions and focused more on the present in their tweets than their news articles.
Expectedly, tweets used more authentic language, fillers (um, err, I mean, you know) and netspeak; resulting in a more casual conversation style and informal rhetoric
A strong anchoring bias effect was evident, such that journalists who had covered previous elections (e.g. 2012) were far more likely to mention these in the current 2016 election reporting

Discussion

Summarising key findings:

“relative to their work in papers, reporters were significantly more likely to use language that was emotional and focused on the present when they were on Twitter. Journalists’ language in tweets also contained more certainty but included fewer analytical words and fewer numerical terms than their news articles, suggesting self-validating and intuitive reasoning” (p17).

System 1 thinking was common in Twitter posts and “especially amplified in journalists’ minds when they navigate Twitter to engage with their audiences in a fast and personalized manner” (p18).

Practical implications is that research highlights that reporters are under increasing time pressure due to the non-stop social media world and this results in “fewer sources, conducted less cross-checking, and were dependent on public relations and politicians for news sources” (p18).

I think it’s also applicable to incident investigation processes (and other processes).

We probably don’t Tweet often about incidents (although software engineering seems to be more visible in this space), but the time pressures to quickly resolve investigations undoubtedly result in a change of thinking.

E.g. pressures result in more use of heuristics, assumptions and selective data sources compared to being able to rely on upward counterfactuals, alternative hypotheses and other useful narratives (link in comments to a study on upward/downward counterfactuals in safety).

Moreover, this adds to a host of other factors that shift the focus and “findings” of an investigation (another link below to that study).

Authors: J Lee, JT Hamilton – Plos one, 2022