The Big Question:
How are algorithms shifting power?
- In society at large?
- Among specific categories of people?
- Within journalism and between the profession and other key social institutions, such as Big Tech?
Ask Yourself:
Does my reporting spread unrealistic expectations (whether good or bad) about AI?
Beware of reporting that adapts either a utopian or dystopian understanding of what AI can do (boomers versus doomers). Focus instead on what systems actually do (or fail to do) and the “guardrails” necessary to make their use safe and equitable for everyone.
What can this algorithm do?
- What can this algorithm do?
- How does this algorithm work?
- What are the challenges posed by this particular use of AI?
- What guardrails are necessary to make its use safe and equitable?
- Is the system only being used for the specific purpose for which it was designed?
- How do people perceive this algorithm?
- Is its operation invisible or unclear to the people affected by it?
How was this system Developed?
- What data does the system use for training?
- Who gathered this data?
- Is the data used appropriate for this purpose?
- Has the data been audited to assure that it does not reflect discriminatory practices such as racial profiling?
What type of content does this system amplify? What type of content does this system marginalize or block?
How Accurate is this System?
- Under what conditions does it make mistakes?
- How often does this happen?
What happens when the system makes mistakes?
- How are system users trained to recognize and resolve errors?
- What reporting processes assure that the public is informed?
- Is there community oversight in monitoring errors and outcomes?
- What protections are there for whistleblowers?
- What are the threats of false positives and false negatives (For more on false positives and negatives, and why the difference matters, see How Journalists Can Systematically Critique Algorithms, by Daniel Trielli and Nicholas Diakopoulos.