Posted:
by QW+ staff writer
Written interviews: 4 ways NOT to use AI and 4 ways it can help
Across all spaces, journalists are telling QW+ the same thing: that they’re fed up with experts answering their email questions in a voice that sounds suspiciously like AI.
They remind us that they’ll approach a QW+ expert because they trust that expert’s judgement. That they are interested in the unique perspectives they can bring on stories, which includes the expert’s lived understanding of the issue on the table.
And when experts use AI to give them a soundbite or key message, they weaken own voices.
Here are four of the most critical ways you lose ground when you use AI to answer interview questions.
Problem 1: AI has its own voice, and it’s definitely not yours
From Claude to ChatGPT, there is a uniformity in an AI-derived voice. It sounds considered, intelligent, and polished. And universally agreeable. But it is a tool that predicts language patterns. It cannot build an argument that shows how a real, complicated human thinks through a problem. Like tea with just too much sugar, the first sip might seem okay, but by the time you’re a third of the way through, you’ve had enough.
In AI-written text, your argument gets lost under a swirl of words that look impressive, but do not carry meaning.
Journalists chase both facts and deeper understanding, based on what people tell them. If what write does not reflect your personal approach to thinking, then what you say becomes useless for their purpose.
Problem 2: AI cannot move beyond mainstream framing
AI is trained within the context of Western-centric, English-centric, Christian-centric, middle-class centric power constructs.
And while its ultra-smooth language language can sound really impressive, it can – and will – slightly shift your language away from your unique perspectives and towards safer, more familiar narratives. With AI writing, we all end up sounding like PR firms from the midwest USA.
Problem 3: You sound credible but not substantive
Problem 2 leads directly into problem 3: that AI’s favourite trick is to simplify and strip out context. The Large Language Models on which tools like ChatGPT are based cannot hold complexity and cannot work within the parameters that one lens might be critical at one stage, and a different one at another. Unless you actively hold them in focus, AI will strip out the complexity that gender, race, class, ability, geography, religion, and so many other factors bring.
Problem 4: What you’re quoted on won’t be what matters most
AI writes in soundbites, and journalists hone into these. Under pressure, a journalist might lift the first “smart” sentence you write, without taking the time to understand the full perspective you bring.
Four ways to use AI instead
One: Use AI to help you understand why this story is important
AI can help you understand the journalist, their audience, and what their interest is in the matter you’re being asked about.
Ask your favourite AI to tell you what they know about the journalist. Ask it to tell you what they know about their readers/listeners/audience. And ask it to bring you up to speed with other news or social media stories circulating currently on the topic you’re being interviewed on.
When you have all of this info banked, ask yourself: in the context of your own area of expertise, what does that particular journalist need their particular audience to understand? Write that down, then ask your AI the same thing. Compare answers. Tell AI why you came up with your answer, and ask it why it came up with the one it did. This is valuable context that can make for a deeper, more contextualised, interview.
Two: Use AI to better understand: “Why me, for this story?”
The same process applies for step two. Once you’ve answered this question yourself, pose it to AI, and discuss your different perspectives. We all have blindspots and AI can help you see yourself through the journalist’s eyes. It can help you understand what perspectives they are looking for and the unique angles that you bring that advance the story in important ways.
Three: Use AI to dive deeply into the one core point you need your audience to understand
If you start with your own thinking, AI can tighten your words to a key message you can stand behind. It can suggest ways to weave that point into repeated answers so that it doesn’t get lost. And suggest questions you can ask back to the journalist to check that what you’ve tried to get across, has landed with them.
Four: Use AI to check what you might be missing
Once you’ve worked through your interview, remind AI of who your audience is before feeding your final Q&A into your tool. Ask it: for this audience, what context is absent? Have we covered the key issues from the perspective of this particular readership or are there gaps we need to fill in before sending a response?
Bottom line: it’s perfectly acceptable to use AI to prepare your contextualised thinking. But you need to write up your answers as yourself. Because at the end of the day, it’s your opinion that the journalist came for. And your opinion – not AI’s regurgitation – that matters for a better world.
Picture: Kit (formerly ConvertKit), Unsplash