A seasonal despondency in journalism
My friend Dhiya has been writing and publishing one blog post every week for the last 87 weeks in a series she simply calls ‘dilettante’. The eighty-fifth edition focused on an issue I am currently engaged deeply with at work: as a journalist trying to make their team’s work digital-first and as a journalist contending with the consequences of their work via the internet. Excerpt:
Pollsters have always known that it is impossible to survey everyone. The entire discipline is built on the idea that a representative sample can tell you something useful about the whole electorate. But what happens when the people willing to answer surveys stop looking like the electorate itself?
In 2026, the people most likely to respond to political polls tend to be more politically engaged, more interested in current affairs, and often highly educated. That’s all great to know, but that’s one set of voters. Pollsters know this, so they compensate, they accommodate for "non-response bias." They weigh the responses, build models, make assumptions about who is missing, and how those missing voices might behave. The quality of modelling is most of the business, it seems. In essence, polling is not measuring exact opinion. They are measuring some opinions and estimating the rest.
That thought struck me because I think we are seeing something similar in audience metrics around storytelling.
For most of my career, the answer to the question "Did this work?" lived inside an analytics dashboard. The internet meant audiences were finally quantifiable, to a very specific degree. We could count clicks, impressions, page views, followers, subscribers, opens, engagements and watch time. Unlike television ratings or newspaper circulation figures, digital media was precise. We could know exactly where they lived, when they engaged, what else they liked. We weren't estimating any more. We were measuring. Entire media companies and several tech companies were built on this ability.
A page view was a fairly straightforward thing. Someone arrived at a page. A number went up. Today, that same page may be summarised by an LLM that speaks to a user. It may appear inside an AI overview. It may be screenshotted and shared by text message. A link may be dropped into a group, and spark a conversation among people who never actually click it. A clip from a podcast might get shared. Forget the hypothetical, I did all of those things in the last 24 hours. It’s really how media is being consumed today.
The most obvious challenge is that audiences themselves are becoming harder to observe. Increasingly, people spend time in encrypted group chats, newsletters, communities, recommendation feeds and AI interfaces. Their behaviour leaves fewer visible traces behind.
But there's another problem. Even when we can observe behaviour, we're often observing the behaviour of the easiest people to measure. The people leaving the clearest traces are not representative of all the people you're trying to reach. For pollsters, it's highly engaged voters who answer surveys. For communicators, it's the people who click, comment, subscribe and engage publicly.
An executive citing a story in a leadership meeting doesn’t appear in monthly engagement reports. A strategy conversation that references a story six months after publication won't show up in analytics. A customer changing their perception of the brand after encountering the same idea three different times is effectively invisible.
My friend is right: interpretation has become a significant part of how we understand our audiences online. I used to think it was something of a paradox: that the internet empowered us to ‘see’ audience behaviour more clearly while increasingly obscuring the audience itself. Thus we had more powerful ways to identify and track readers, relative to print journalism, while also increasingly having to interpret that data ‘correctly’. Then I realised it was not a paradox, but something resembling irony, because the internet also created new ways for readers to consume the news, especially many that had nothing to do with the needs of journalists.
I would like to use this passage as a springboard to address a related concern. A colleague of mine suffers from something I call “seasonal despondency”. He is a successful and well-regarded political journalist. But once every few months, he starts to wonder if his work has any impact on the people at large, especially in the face of majoritarianism, nationalism, trenchant xenophobia, and successful disinformation campaigns run by right-wing propagandists. He starts to become despondent about his life’s work, wonders if he is wasting his time, and if he should have studied to become an engineer or have become an activist instead. His mood eventually improves after conversations with friends (including me) or if something in his work manages to distract him.
I think such seasonal despondency is more widespread among journalists than I have evidence for, and it is due in significant — if not in entire — part to the diminishing connection between us journalists (especially those not on the political right) and our readers. Most of the noise we are able to hear more readily, to use Dhiya’s analogy, is made by those prepared to be the loudest in public. But are they representative of the people at large who consume our journalism?
We know that the loudest audience consists of people sufficiently motivated to write: whether to a journalist or to post about it on a different news platform, or to complain, praise, share, quote-tweet, etc. And journalists who challenge majoritarianism et al. encounter a remarkably intense volume of hostile reactions online, and could mistake that for evidence of the public response at large, especially when responses from other quarters are muted or altogether absent.
Readers who find these journalists’ work useful but have no reason to announce that will remain invisible. It is a well-known fact today that such announcements are fraught. If you do, you know you will need to be prepared for characteristically low-quality responses, and often also respondents, on social media platforms and large WhatsApp groups. Many research papers have also documented a negativity bias implicit in social media networks: i.e. they amplify bad news, especially if it is also sensational, more because that is likely to elicit more impassioned responses from users (not unlike the “if it bleeds, it leads” adage of ‘old school’ journalism).
The “seasonal despondency” is really the result of what ought to be recognised as a well-defined psychological trap. Like my colleague, journalists can spend years, even decades, producing work that is discussed and incorporated in the ways in which hundreds of thousands of people understand the world, yet these days, the same journalists are also more exposed to the opposition to that work, especially in the form of abuse and dismissal.
One interpretation that may be useful, especially at journalism organisations like The Hindu, where the volume of audience metrics available is very large, is to track the number of unique users on an article along with the number of sessions and the average time on page. While there being few incentives for readers to speak positively, or at least constructively, about an article visibly somewhere on the internet is a sobering notion, knowing whether people are interested in articles with a particular substance and keep returning to it could help patch that gap.
Of course, as my friend says, the ultimate remedy for the seasonal despondency would be to get on the ground and “talk to people” and “listen to conversations”. Which is not easy to do, for many reasons, but perhaps that is one of the next big problems to solve as our journalism becomes digital-first and, perhaps inevitably, digital-native.
