Building AI into the investigative workflow

Talk presented to the AI in Journalism Education pre-conference, October 9 2024

1. Building AI into the
investigative workflow

Å integrere AI i den undersøkende arbeidsprosessen

2. What I’ll cover
●How AI can lower the barrier to entry in
investigative journalism
●Assessing and addressing the risks

3. The story so far
●MA Data Journalism at BCU: machine learning
used in student investigations since 2017
●2022: ChatGPT means AI becomes
non-investigation-specific (a “gateway drug”)
●Accelerates students’ coding, etc.

4. Bradshaw (2024) AI in
investigative journalism:
mapping the field

5. The problem
●GenAI threatens core journalism values
●GenAI threatens journalism’s status
●GenAI threatens journalists’ jobs

6. Not all our work is about facts.
But…

7. Not all our work is about facts.
Journalism is also about creativity.
But…

8. Not all our work is about facts.
Journalism is also about creativity.
Succinctness and clarity.
But…

9. Not all our work is about facts.
Journalism is also about creativity.
Succinctness and clarity.
Coding.
But…

10. Not all our work is about facts.
Journalism is also about creativity.
Succinctness and clarity.
Coding.
And reflexivity.
But…

11. Research

Scope diverse sources,
explore documents, form
advanced searches, and
write/fix code for scraping
and analysis
2
Pre-production

Idea generation and stimulation:
identify and map systems and
rules, apply brainstorming
frameworks (iceberg model, 5
whys, 8 angles of data journalism).
Planning.
1
Post-production

Optimisation, distribution
and reversioning
4
Production

Identify jargon and bias;
improve spelling, grammar,
structure and brevity
3
Applications of genAI in the journalism process

12. Research

Scope diverse sources,
explore documents, form
advanced searches, and
write/fix code for scraping
and analysis
2
Pre-production

Idea generation and stimulation:
identify and map systems and
rules, apply brainstorming
frameworks (iceberg model, 5
whys, 8 angles of data journalism).
Planning.
1
Post-production

Optimisation and
reversioning
4
Production

Identify jargon and bias;
improve spelling, grammar,
structure and brevity
3
Applications of genAI in the journalism process

13. You are an [investigative journalist]
working for a [national broadcaster]
aimed at an audience aged [20-40].
Identify parts of the [UK school] system
that might be suitable for an
investigation.
Starting with systems:

14. https://onlinejournalismblog.com/2024/07/10/investigative-journalism-an
d-chatgpt-using-generative-ai-for-story-ideas/

15. You are a [UK journalist writing for a
rural audience] looking for [feature]
story ideas relating to [education].
What rules do [UK schools] have to
follow in relation to [school uniform]?
Starting with rules:

16. Why rules?
●Rules having unintended consequences
●Rules not being followed
●Rules not being enforced
●The need for new rules

17. Take idea 2. Apply the [ SCAMPER method]
to it: [suggest variations of the idea
which substitute elements, combine it
with other ideas, adapt, modify, put to
another use, eliminate elements, or
reverse].
Applying a framework

18. https://medium.com/@headlinexplorer/investigative-journalism-iii-hypot
hesis-based-inquiry-557f48ea8dff

19. Omfang
7 vanlige vinkler for datahistorier_
Veksling
Dårlig/åpneRelasjonerUtforske
VariasjonRangering
(+ Saker)
Icons: the Noun Project: Becris (scale), Adrian Coquet (change and ranking), Kirby Wu (variation),
Aradila Studio (explore) Trevor Dsouza (relationships), Iconpai (bad data), Kirill Ulitin (leads)
https://onlinejournalismblog.com/2020/08/11/here-are-the-7-types-of-stories-most-often-found-in-data/

20. A new story reports that [scientists
have confirmed climate change is a major
reason the UK suffered such a
waterlogged winter]. You are a [data
journalist] looking to do a [data-driven
follow-up investigation] on this aimed
at a [UK] audience. Come up with 5 short
pitches for story ideas that you can
present to your editor when you next
meet her.
Injecting knowledge

21. Research

Scope diverse sources,
explore documents, form
advanced searches, and
write/fix code for scraping
and analysis
2
Pre-production

Idea generation and stimulation:
identify and map systems and
rules, apply brainstorming
frameworks (iceberg model, 5
whys, 8 angles of data journalism).
Planning.
1
Post-production

Optimisation and
reversioning
4
Production

Identify jargon and bias;
improve spelling, grammar,
structure and brevity
3
Applications of genAI in the journalism process

22. Research opportunities
●FOI
●Advanced search (‘Google dorks’)
●Scraping (help with code)
●Analysis (help with formulae, cleaning)
●Document sets (search across)
●Diverse sourcing

23. Story stage Opportunities Risks
Idea generation 1.Little time for idea exploration
2.Low exposure to potential inspiration.
3.Limited usage of models such as SCAMPER, systems thinking etc.

GenAI can increase exposure and reduce time required. Prompts can
require application of models.
Hallucination: low (ideas can be easily
rejected)
Knowledge cutoff: medium (new information
may need to be injected)
Bias: high (prompts must specify diversity)
Environmental: high (results can be achieved
without genAI)
Story research 1.Research lacks diversity
2.Lack of time for document exploration

GenAI can increase diversity and assist with technical barriers.
Hallucination: medium (material must be
checked)
Knowledge cutoff: medium (new information
may need to be injected)
Bias: high (prompts must specify diversity)
Environmental: medium (genAI required for
some tasks)
1.Lack of literacy/confidence in FOI, scraping and advanced search
2.Data needs cleaning

Hallucination: low (easily testable)
Knowledge cutoff: low (not reliant on up to
date info)
Bias: low
Environmental: high (results could be
achieved without genAI, but would require
training)
Opportunities vs risks

24. Research

Scope diverse sources,
explore documents, form
advanced searches, and
write/fix code for scraping
and analysis
2
Pre-production

Idea generation and stimulation:
identify and map systems and
rules, apply brainstorming
frameworks (iceberg model, 5
whys, 8 angles of data journalism).
Planning.
1
Post-production

Optimisation, distribution
and reversioning
4
Production

Identify jargon and bias;
improve spelling, grammar,
structure and brevity
3
Applications of genAI in the journalism process

25. You are an experienced editor who hates
it when a journalist gets too close to a
story and ‘overwrites’ it - that is,
uses too many words, or sentences that
are redundant. You also hate jargon and
prefer clear language that is accessible
to a normal reader.

Identify any potential jargon or
overwriting in this story and the steps
the journalist can take to address
those.

26. https://onlinejournalismblog.com/2024/09/26/using-generative-ai-as-a-s
preadsheet-and-data-cleaning-assistant/

28. Your publication has just introduced a
policy of ensuring that all stories
reflect the diversity of the communities
it seeks to serve. Those communities
include people of different faiths,
social class, ethnicity, gender, and
sexuality. It includes both rural and
urban readers, old and young, and people
with disabilities.

Identify any potential bias in this
article and steps the journalist can
take to address those.

29. https://onlinejournalismblog.com/2024/10/10/identifying-bias-in-your-writ
ing-with-generative-ai/

30. https://onlinejournalismblog.com/2024/10/10/identifying-bias-in-your-writ
ing-with-generative-ai/

31. Key points
●Opportunities are highest, and risk lowest, in the
earlier stages of investigations
●GenAI works best to extend thinking beyond
limitations of experience and routine
●Risk assessment needed, including
environmental impact of usage