Testing tools and AI - ideas what to try with some tool examples
Effects revised
1. The effects of source
documents on recall
and credibility
Feb. 28, 2012
Megan Duncan
Master’s thesis
2. Lit: Role of journalism
• Human
Sources • Documents
• Simplify complex ideas
Interpret • Put new ideas in context
• Traditional: Inverted pyramid
Report • Newer: Narrative
5. Hypotheses and Research Questions
Recall
• H1: Exposure to a copy of a court document and a traditional
news story about that document will increase the reader’s
recall of the story when compared to being exposed to a news
story alone.
• RQ1: What factors influence recall?
Credibility
• H2: Exposure to a copy of a court document and a traditional
news story about that document will increase the reader’s
perception of credibility of the story when compared to being
exposed to a news story alone.
• RQ1: What factors influence credibility?
6. Participants in the experiment
Days the online Condition
experiment was
open to participants
77 Treatment
81
Completed the Control
experiment
Removed for missing data
The
Gender
Removed for prior mean
knowledge age was
29 Female
71 Male
Used in analysis
7. Results: Open-ended summaries
“A Jewish man from an American
“Someone from Boston offered to company was selling information to
sell a foreign country secrets from another country, Country X. The
FBI caught him by setting up a sting
his company.” operation.”
Participant 174, control Participant 131, control
“Doxer worked for an Internet
“A man named Doxer was technology company in Massachusetts
accused of committing a crime. and offered to sell an undisclosed country
information about confidential matters.
He claims he did nothing The country got wind of it, told the
wrong but the authorities, and United States, and then worked with the
the criminal report, say FBI to help build a case against Doxer.”
differently. His mother is also Participant 27, treatment
a ‘terrible person.’”
Participant 66, treatment
8. Results: Recall
Participants were asked recall questions about a news story
Those who saw the source Those who saw ONLY the news story
document answered on average answered on average
questions correctly. questions correctly.
When participants who answered questions correctly were removed from
analysis, those who saw the source document answered percent correctly.
Those who saw the news story only answered percent correctly.
Independent Sample T-test of Proportion of Recall Questions Correct by Condition
Mean proportion SD t p
Control (77) .7217 .26412 -1.652 .101
Treatment (81) .7866 .22714
Participants who answered all recall questions incorrectly removed (N=4)
Control (75) .7410 .23915 -1.872 .063
Treatment (79) .8065 .19144
9. Results: Recall
questions had significance
between conditions
What did the person charged with a
Which agency investigated the crime?
crime leave at the “dead drop”?
Treatment Treatment
Correct Correct
Incorrect Incorrect
Control Control
Mean SD t p Mean SD t p
Control (77) .73 .448 Control (77) .58 .496
Treatment (81) .85 .357 Treatment (81) .84 .369
-1.925 .056 -3.652 .000
10. Results: Credibility
Participants were asked credibility questions
questions had statistical difference between control and treatment
On a scale of to , participants were asked to rate their perception of the news article.
The average credibility score was . The average trust score was about
the same: . Overall, participants rated it as more credible than not.
Independent sample T-test of perception of credibility by condition
Mean SD t p
Control (77) 29.1429 5.89539 .235 .814
Treatment (81) 28.9012 6.92929
11. Results: Hypotheses
• H1: Exposure to a copy of a • H2: Exposure to a copy of a
court document and a court document and a
traditional news story traditional news story about
about that document will that document will increase
increase the reader’s recall the reader’s perception of
credibility of the story when
Recall of the story when
compared to being exposed
to a news story alone.
Credibility compared to being exposed
to a news story alone.
• Based on these results, • Based on these results, there
there was not enough was no evidence to support
Not evidence to support H1.
Not H2.
supported supported
12. Results: Research questions
Race/ethnicity News
Time p=.070
p=.023
Interest
p=.014
Story
Interest
Recall p=.000, Credibility p=.007
Interaction:
Condition X Story Time
Interest p=.016
p=.007
13. Discussion: Recall
Mean proportion of recall questions answered
correctly by condition influenced by story interest
interest
Recall
Story
Among those who
answered at least recall
question correctly, there
was statistical difference
between conditions. This
may indicate an effect of
source documents on HSM
those who actually read had an effect on recall and not on
the news story. credibility. This is the opposite of what
the literature predicted.
14. Discussion: Credibility
Providing an official document that
When participants were
carried with it the names of federal
presented a story written in
courts, a federal judge, and a FBI
inverted-pyramid style but
agent didn’t improve that credibility.
lacking any association with a
news organization or
particular author, they
reported that they believed
in and trusted in the
information there. While the use of anonymous sources
harms the credibility of news,
providing extensive details about the
source didn’t improve credibility.
15. Discussion: Future research?
Are certain types of
Other types
info effected more?
of source
documents?
How much
source
information is
Who chooses to
needed to be
read source
credible?
documents?
Why/when does Is the way news
the news provide credibility is judged
source changing among
documents? young people?