Editor’s note: Deepnews.ai is a technology company, though we also like to look at conversations happening around the use of algorithms to highlight information online. This is one of the occasional posts we do speaking to someone with something to say on topics that we find interesting.
By Christopher Brennan
Speaking personally, one of my favorite aspects of what Deepnews.ai does is that it highlights stories from smaller sources, particularly local news.
As someone who has worked at a local paper in the U.S., I enjoy seeing the good work being done from metro outlets that have been serving their communities for decades, as well as upstart non-profits that are trying to fill in gaps in coverage. Oftentimes a well-done local story, whether from Corpus Christi or Concord, NH, can have national or international resonance.
The economic decline of those local newspapers is an issue that is always on my mind, but my interest was piqued again when I saw a paper in Nature from Sean Fischer, a researcher at the University of Pennsylvania, along with Dr. Kokil Jaidka at the National University of Singapore and Penn colleague Yphtach Lelkes.
In it they discuss the placement of local news stories in Google News results, as well as factors that may explain why they never rank very highly. I gave Fischer and Jaidka a call this week to discuss their work. Here is a transcript of our conversation lightly edited for length.
Brennan: So to start off, how did this project on local news come about? You’ve both worked on other research involving different aspects of the digital space, but why local news?
Sean Fischer: So when we started off, local news wasn’t necessarily our intended topic. We were interested a little bit more in political bias in the search results. We were interested in basically the mechanism of “Does where you are affect the news that’s returned to you, but in a partisan sort of way?” And the answer there is a really boring “no” because pretty much everybody sees exactly the same news outlets. But Kokil had put a lot of time into getting this really great dataset. And so we said, you know, is there anything else that in the realm of bias and attention that would be valuable to test and to study?
This was in 2018, so Dan Hopkins’s book had just come out on the nationalization of local politics. And you were hearing in that 2018 election cycle about how much Trump and everything was affecting otherwise local races. And we said, you know, maybe there’s something to that. These local outlets do things that are the base of our local governments, local communities. Are they being supported? Because if not we know they’re already in crisis and that’s really well documented. And we know they’re in crisis because they don’t have people going to their websites, basically. The problem can sort of be reduced down to that. And so we said, you know Google News would be really interesting to see. Is it helping or is it hurting or is it somewhere in between? So we expanded the data collection and recircled around this idea of where. Where is the local news on Google News? And obviously we found that it’s not anywhere helpful, mostly.
Brennan: So the headline is “local news displaced by national news,” essentially. But you also looked into matters beyond that as well.
Fischer: So the results are contextualized in a few ways. The first is that what you search for does matter. So when we did this in 2018, Donald Trump was threatening to shut down the government and then he sort of went through with it at the start of 2019. And so, if you searched for “shutdown” which was a big national topic, you got a lot of national news. And you got a lot of national news even deep into the search results. They were the outlets you got.
But if you searched for, say, “school board” or “traffic” or “weather,” things like that, the results would have more local outlets in them. So that’s a big point here is that if you’re searching for something that is a local or a more locally oriented topic, you’re going to get more local outlets.
However, what we find is that even for local topics, if you look at the really valuable search results spots, those first 5, 10, 15-ish spots, what used to be counted as a single page before Google moved to infinite scrolling, the share of those spots that go to local outlets drops as you get closer and closer to the top. Even for local search terms. For a lot of our local search terms, you’d be going along smoothly, smoothly, smoothly, but as you got into that top area, it just dropped right off.
So finally we circled back and we said OK. So how does where you live matter in this process? And we looked at that and we said does the probability of seeing a local outlet change if you live someplace where there’s a healthier local news economy, all of these predictors of local news health were used and what we found, that’s not the case. It doesn’t really matter, the health of your local news economy.
So we sort of spun it back round and we said “So wait. Is this just a rich get richer sort of phenomenon?” And we remodeled. The number of times an outlet gets returned, is it based basically on its circulation, how many subscribers it already has for legacy print media outlets? And that’s what we did find that was a very clear pattern. That the rich are getting richer. Outlets that already have a lot of subscribers, well, they’re doing fine and Google is pushing them back into the top.
So it’s not really disrupting the sort of economic inequality in the media marketplace as it is. And we finish off the paper. If you stop there that doesn’t sound super positive. That doesn’t sound super good. But obviously, the effects of Google News or any platform depend on how the users actually use the platform. And while we didn’t observe anybody using the platform directly, we didn’t pay some people to install some software so that we can track their searches or something like that. But what we did do was we collected the Google Trends data for Google News for the terms we looked at, the 32 terms. And what we found was that there was a small but consistent and significant increase in interest for local terms compared to the national terms or the general terms that we searched for.
So we sort of took that to be: OK, Google News is not structured in a way that is necessarily automatically helpful for local outlets or smaller outlets, or new outlets. Really any outlet that doesn’t have an established following. But user behavior is such from what we can measure passively that the platform could be supportive of those outlets. That people aren’t only going to Google News and looking for news about the latest scandal in Washington, or the protests in Portland, or the wildfires or a hurricane when it comes and hits. People are actually going there and they’re interested in what’s happening in their community. There seems to be evidence that they are using the portal as a way to learn about what’s happening around them, they’re not just using it as a sort of a quick summary of the nation on any given day.
Brennan: Not to give Google any advice, but looking at this, is what Google offers incongruent with what people want? I mean, ostensibly the results that Google are showing are the most relevant results, but also what people want.
Dr. Kokil Jaidka: I think the results can be taken to mean that the people who are deciding how the nation interprets any news topic, is the national news, not the local news. Which is a pity when you’re talking about local issues. We found similar results when we were looking at how the Portland protests were being discussed. So the first thing is, this does have implications for who’s setting the frame, and who gets to be on the front page and who doesn’t.
The second is, Google did take cognizance of our research. And they did say that they are now making a conscious effort to include more local news and their reporting. So the initial data was collected last year and even this year when we collected more data, we did notice just a little bit of difference in how our search queries are being interpreted. So first of all, Google did take cognizance of the fact that having local news being reflected in the top results is something and it’s important and why it’s important.
Secondly, the fact that people are going to Google and searching for local articles suggests that, first of all, Google’s default settings on Google’s homepage aren”t providing that information. As Sean said, it’s all set up to encourage you to find out about your country, not necessarily about your community. And that could be your signal. That could be another signal that Google could consider. It could consider adding a separate tab or tiny panel which provides news on your community. I think it is very meaningful that the search results, the traffic for different queries is going one way and Google’s own priorities, Google’s own list of things on its homepage and ordering of news results is going another way.
Fischer: I think it’s worth thinking about as well is when people are asked about what types of outlets they trust or that they prefer to turn to. The Pew Research Center is always surveying on this. Local news always does better than big branded national news outlets. Especially right now, when you have multiple elected officials questioning the high profile outlets: CNN, New York Times, Washington Post. They’re always under fire from somebody at this point. And honestly, across parties at times.
So people are queued up to really want to listen to and trust in local news for more than just the topics of local interest. Local news outlets covering national news appear to be more trustworthy to folks. The second is that Google is actively working in this space. In the last week or week and a half at this point, they’ve announced they’re giving out a billion dollars over the next few years in partnership with outlets. It’s thinking about ways that they can add and promote content on their platform, specifically on Google News.
Brennan: It’s a limited number of outlets to begin with though.
Fischer: Yes, that’s very true and they’ve picked some odd partners in a variety of countries. I think the Nieman Lab position on it, that this is just to stave off having to pay out billions in regulation later, is probably perfectly fair. But it is clear that they’re willing to at times think about how can we modify the platform. That they don’t see the platform as stable at all.
Jaidka: And adding to the number of ways in which they’re trying to support the news economy, they have also announced a new licensing program to support the news industry. They say that this is to provide an even playing field to smaller news outlets. Again, one other way in which they’re creating or trying to increase the viability of these smaller outlets.
And speaking of viability and economy, it is important to talk about the business aspect of this whole problem. It’s not just useful to the society, in terms of having more local news in their search results, but it’s also important for the economy because right now, everything is working off of clicks and traffic. So it’s not simply a question of having a representative voice, but it’s also a question of how they’re gatekeeping the revenue of all these smaller outlets. And then it becomes a vicious cycle because if they’re not getting the money then they will find other ways to support themselves. They will turn to paying their journalists poorly, giving them worse hours to work, giving them other stories to write, outsourcing, or basically collecting, rather than sending out their own reporters. Just using Associated Press or Reuters, rather than sending someone to the ground.
Brennan: You mentioned in your paper in passing the idea that a large amount of original reporting out there is coming from these local news outlets that aren’t particularly well represented. I dont think you did any sort of content analysis in this paper but right after you stopped collecting data for this paper, Google News did change its algorithm to focus more on what it said was was original content. I was wondering, especially if you’re doing more research now, whether you’ve seen that.
Jaidka: That’s a great point. In fact, Sean got some comments about this as well some feedback. You want to take this?
Fischer: I mean it is a feature of right when the data was stopped that we didn’t get to look at that specifically. So I think we sort of see it as the next big frontier for this type of research. Our reviewers, and people at several different conferences where this got workshopped and presented, felt that this was the next big leap. That it’s important to describe and measure the problem at this level that we’ve done, but then to get into it.
I think out of Duke, out of their one of their journalism centers, it was one of the first big flashy reports of, “Hey, you know, even like the local news outlets that do exist in a variety of different forms, whether they’re newspapers or local television websites where they’re posting short stories. How much of that content is even meeting the needs of the community?” Communities have a certain priority. There’s certain information they really need to have access to. Are [local outlets] even doing it? And so there is work out there now saying, “Let’s look and say, OK, not all local news is helpful.” To start to qualify it, and that’s where these content analyses are coming in.
Jaidka: To add to what Sean saying, honestly, it is something we want to do, and it’s definitely my expertise because I do a lot of language stuff. We felt it was important to first understand the claim. And then we have suspected that, because there are a few conglomerates who own most of these local news outlets anyway. Secondly because of the lack of the funds to pay local journalists they might be sourcing stuff from Reuters. We did collect these things for the data that we have. It’s just that we haven’t looked at it yet, because of wanting to do a robust enough job on the first paper.
Fischer: There’s another one beyond the sort of what they call these “ghost newspapers” that Kokil was describing, where it’s one person and then it’s all AP or Reuters.
So the way we did our searches we searched everything and said “near here” and set our parameters in the browser. And then we collected the results, but then all we did was we tagged to the outlets as being local or not local. We haven’t actually worked out, are the stories from local places about the place? Are they even covering the area we searched?
So if we searched “Philadelphia” are we getting stories from the Inquirer, or are we getting stories from random local news in some other state, even? So there’s a lot to be gleaned really from getting into the actual text of the articles and coming up with a good pipeline for getting that text from the data. And matching it back to local news.
Brennan: At Deepnews we are of course, all about the content. You also touched on polarization a few moments ago, which is a hot topic. If that was something you were looking for at the beginning of the project is that something you would return to in terms of the impact of minimal local news on on polarization?
Fischer: Yeah, it’s a very hot topic, It’s a good point. I mean there’s always a possibility to go back to it. Our approach was to sort of look at … We had one of these data sets where the domains have all been coded somewhere on the spectrum. And we looked at those. And what we found was at that level what you were seeing wasn’t polarized, but you’re right that the content matters.
It’s taken as a settled point. There was a great paper from Josh Darr and Johanna Dunaway, and I’m forgetting the third co-author, last year, or the year before, that really did show that just when you lose the local news, your [local] elections just become straight ticket, party-based. Everybody uses national cues. Something we’ve thought about and talked about is trying to disentangle the mechanism of that. Because it’s an agreed upon point that it happens. Something we’ve talked about is “Why is that the case?” If you replace local media with national media, what’s the cueing process? Is it just references to the parties and things like that? So that’s a content level. That can maybe help us expand our understanding. At a time of interest I think everybody sort of looks around and goes, “Yeah, it is polarized.”
Brennan: This can be the last question. In a lot of the discussions around fixing things like misinformation and news in general there is a split of focusing on individual user behavior versus the platforms. And I was wondering, because your research focuses pretty squarely on the platforms here, what’s your take on whether it is possible to to change the decline of local news or the economic pressures on it without looking at the platforms.
Fischer: That’s a really good question. Here in the United States, again there is something that has cropped up in the last 10 days. Congress is sort of beginning to move on the language that it has had all pandemic about, “We need to be thinking about local news” and “We need to take this opportunity to protect and save local news.”
In the last week or so a group of senators proposed: why don’t we make a sort of bipartisan or non-partisan elite, expert commission. And they’re going to study the problem and tell us you know give us recommendations for what we should do. If you bump around in these different circles of CJR and Nieman Lab and stuff everybody sort of has their take. Should they just make a big slush fund and create a big fund that all the local news people can tap into? And we’ll fund everybody forever. And yeah if they want to that’d be great.
But correct me if I’m wrong, but I think our consensus is that any practical solution, probably needs to include thinking about multiple digital platforms: Google, Google News, Facebook, Twitter. These are sites that are pretty well documented as pathways to news and regulations or any intervention to save local news probably needs to think directly about where people get it. And that’s these platforms.
Jaidka: I mean [local news] could be putting out quality content and no one would know if the platform didn’t want anyone to know. Again, there’s enough evidence that, whether intentionally or unintentionally, polarizing news, or triggering news, or clickbait is amplified on social media platforms. I don’t think the criteria is the same for local news but definitely the way queries are interpreted may not be appropriate for the local context, unless there is some stuff being set up at the back end. Which goes to say, I believe no solution can be agnostic to the role that platforms play.
Very importantly, in talking about misinformation disinformation I think one thing that Spotify and Apple did was to stop listing Infowars completely. That’s a decisive step that they took in order to control misinformation. It of course would have cost them some revenue. It did drop visits and listens to Infowars, obviously, which is a great thing. That may have been the one thing that would that actually hit the Achilles heel, which is hurting them, where the money’s coming in from making it financially not viable to put out misinformation. On the other hand there is making it financially viable for local news to continue and to find quality content. Something that currently isn’t possible.