From Radio to Twitter: How the Tone of News Twists Democracy’s Arm

In the summer of 1979, local television anchors around the world were still delivering the news in their best Walter Cronkite timbre. A handful of major newspapers governed the national conversation, radio talk shows stoked the occasional controversy, and cable news was only beginning to nibble at broadcast’s hegemony. GDELT’s massive data on global media from that era reveals an intriguing pattern: the overall tone of news coverage (negativity versus positivity) and the measure of conflict/cooperation (what GDELT calls the Goldstein scale) formed a modest, upward slope. True, news could be dour, but it was generally tempered by a sense of equilibrium—fewer spikes in rancor, fewer lurches into extreme conflict.

Then, look at the same GDELT graph from 2012 onward, and it’s like stepping into a different universe. The gentle slope has turned into a sharp diagonal shift, with negativity skyrocketing and conflict drawing ever closer to the headline’s center stage. In a sense, that single graph, color-coded with a swirl of data points, tells a story as loud as any front-page editorial. After social networks arrived as the new Goliaths, the equilibrium faltered. Our shared conversation—in the big blob of aggregated media—got harsher, gloomier, and far more polarized.

---

The Before: Radio, TV, and National Newspapers

We’ve seen upheavals in political communication before. In the 1930s, radio freed politicians from the constraints of print, prompting Franklin D. Roosevelt to pioneer his “Fireside Chats,” while the Nazis, as Adena and colleagues showed, weaponized broadcast to considerable effect. Television’s spread in the 1950s and 1960s added new spectacle to campaigning: candidates had to perform visually, not just verbally. Journalistic giants like CBS or the BBC kept a gatekeeping role, deciding who got to speak and when. Meanwhile, national newspapers—from the New York Times to Le Monde—dominated the daily political discourse, delivering columns that reached millions. Scholars like Matthew Gentzkow connected TV’s market penetration to voting behavior, while others pointed out that big papers replaced local news, reducing coverage of community-level politics.

Yet for all the swirling transformations, these legacy media platforms still had guardrails. Government regulation of radio content was standard, especially to prevent extremist or manipulative broadcasts. Television likewise abided by fairness doctrines, to varying degrees. Even large newspapers had editorial boards that (theoretically) strove for balanced coverage.

---

The After: How Social Networks Changed the Script

Then came social networks. Facebook took root in the mid-2000s, but it was around 2012—observing both GDELT’s time series and user adoption curves—when platforms like Twitter, YouTube, and a blossoming Facebook ad machine truly merged with mainstream politics. It wasn’t just a new medium; it was a new means of measuring public sentiment. Studies by Tumasjan and Gayo-Avello tried to predict elections from tweets. Others, like the methodology developed by Cárdenas-Sánchez et al. (in which I am a collaborator), mapped how political mention networks converge—or don’t converge—on specific candidates.

One alarming takeaway from these mention-network analyses is that the second largest eigenvalue (a fancy measure of how factions form and resist merging) is climbing in many elections around the globe. Or, as a researcher might put it: convergence is slowing. People aren’t gravitating toward a single conversation. Instead, they’re clustering into echo chambers, each retweeting its own “side,” while negativity soars. GDELT’s swirl of data, with negativity ramping up and conflict ticking higher, lines up almost too perfectly with that narrative.

---

The Main Finding: A Diagonal Lurch into Conflict

That shift—depicted starkly in the graph many of us have stared at in equal parts awe and dread—shows how post-2012 news coverage and commentary effectively bend the curve. Before, negativity rose modestly in tandem with conflict, forming a moderate slope. After 2012, you see a diagonal leap—almost like the data points themselves decided to stage a revolution. Tonally, we have become more cynical; conflict is more prominent, and moments of measured, nuanced debate appear lost in the churn.

What explains this diagonal? In part, it’s the rise of what some call “click-bait democracy.” Freed from the old editorial filters, social networks amplify whichever posts trigger the biggest emotional response. Add to that a new generation of politicians who harness outrage as a brand. The synergy—fueled by retweets, memes, and the dopamine of notifications—pushes content that’s angrier, more suspicious, more in-your-face. GDELT’s tone index catches that negativity in the broader media too, which often echoes social media’s hottest takes.

---

Democracy at the Crossroads (With a Billion Tweets in Tow)

That’s where the Elon Musk factor arrives, like an eccentric ringmaster strolling into an already frenetic circus. Twitter, once a major but not omnipotent force, is now under an owner who claims to champion “free speech,” even as he selectively reshuffles the platform’s policies. Some worry that Musk’s Twitter could tilt the playing field in subtle ways, restoring suspended accounts, tweaking algorithms, or giving certain political voices special leeway. Then there’s the specter of Donald Trump in 2024—back online, brandishing 280-character blasts that can overshadow more sober, nuanced campaign messages. If negativity is the coin of the realm on social media, one wonders: does the richest man on Earth effectively become the new gatekeeper of democratic discourse?

But it’s not only about a single figure. The broader philosophical question is: Are we losing democracy to virality? In the radio era, we regulated licenses; with TV, we insisted on some standard of broadcast fairness. Now, the balkanization that scholars worried about in the 1990s—where the internet fragments us into micro-publics—feels realized on an enormous scale. And it’s not just a matter of what we read; it’s how we behave. The more conflict and negativity intensify, the more we reflexively see opponents as enemies rather than fellow citizens.

---

A Glimmer of Hope?

Of course, the very graph that depicts our gloom could also be our path to a better understanding. GDELT’s minute-by-minute global analysis, combined with mention-network approaches (like the ones in Cárdenas-Sánchez et al.), reveal patterns quickly—potentially enabling real-time identification of disinformation campaigns or emotional contagions. If enough policymakers, media organizations, and everyday citizens recognize when the tone is tipping over, we might collectively pivot toward healthier discourse.

Yes, that’s a lot to hope for. These patterns might deepen before they break. Radio and television were in turn hailed as unifying and demonized as manipulative. They transformed politics dramatically but didn’t entirely demolish democracy. Perhaps social networks, too, will prove to be a double-edged sword, eventually finding some equilibrium or spawning new norms.

Still, the data in that post-2012 graph is a startling mirror. If we stare at it long enough, we might see that democracy itself is not guaranteed—especially if conflict is the easiest path to a share, and vitriol is the reigning currency. Maybe the best use of this research, from GDELT-based negativity metrics to mention-network second eigenvalues, is to shine a spotlight on our own role. After all, it’s us—the users—scrolling and sharing the content. And maybe, just maybe, if enough of us decide that a more constructive conversation is worth having, we can realign the slope of that diagonal, inch by inch, back toward cooperation.

---

Some references:

• Adena, M., et al. “Radio and the Rise of the Nazis in Prewar Germany.” The Quarterly Journal of Economics, 130(4), 1885–1939, 2015.
• Brown, R. J. Manipulating the Ether: The Power of Broadcast Radio in Thirties America. McFarland, 2004.
• Cárdenas-Sánchez, D., Sampayo, A. M., Rodríguez-Prieto, M., & Feged-Rivadeneira, A. “A comparative framework to analyze convergence on Twitter electoral conversations.” Scientific Reports, 12, 19062 (2022).
• DeGroot, M. H. “Reaching a consensus.” Journal of the American Statistical Association, 69(345), 118–121, 1974.
• Gentzkow, M. “Television and voter turnout.” The Quarterly Journal of Economics, 121(3), 931–972, 2006.
• Gayo-Avello, D. “A meta-analysis of state-of-the-art electoral prediction from Twitter data.” Social Science Computer Review, 31(6), 649–679, 2013.
• Norris, P., et al. A virtuous circle: Political communications in postindustrial societies. Cambridge University Press, 2000.
• Tumasjan, A., et al. “Predicting elections with Twitter: What 140 characters reveal about political sentiment.” ICWSM, 2010.
• Van Alstyne, M. & Brynjolfsson, E. “Could the Internet balkanize science?” Science, 274(5292), 1479–1480 (1996).