Tag: Online Discussion

Posted on by

Sockpuppets that agree with their operator inside the same thread

One person can win an argument much faster when they bring three imaginary friends.

That is the simplest form of sockpuppetry.

An operator controls multiple accounts and lets the audience believe they represent independent people.

Account A makes the claim.

Account B agrees enthusiastically.

Account C answers a critic.

Account D upvotes the whole performance.

The visible thread now contains a crowd.

The actual number of independent participants may still be one.

False independence is the product

Multiple accounts are not automatically abusive.

People keep separate professional and personal identities. Moderators use test accounts. Developers run bots. Writers use pseudonyms.

The manipulation begins when the operator uses the separation to obtain something that depends on independent people.

Stack Overflow’s long-standing community guidance expresses the principle clearly: multiple accounts are allowed, but using them to vote for one another, answer one another, approve one another’s edits, or otherwise gain benefits unavailable to a single account is treated as sockpuppet abuse. See Meta Stack Overflow’s summary of the multiple-account rules.

Social platforms have documented the same broader behavior at larger scale. Meta’s coordinated-inauthentic-behavior reports repeatedly describe fake accounts posting, commenting on, and liking their own network’s content to make it appear more popular or independently supported than it was.

Agreement is more persuasive when it appears independent

The second account matters because it changes the social evidence.

A seller saying my product is excellent is advertising.

A stranger saying I bought this too and mine was excellent looks like corroboration.

If both accounts belong to the seller, no second witness exists.

The same trick works in arguments. A sockpuppet can ask a convenient question, praise an answer, attack an opponent, or create the impression that several ordinary readers reached the same conclusion.

Similar opinions are not proof of shared control

This is where lazy accusation becomes dangerous.

Two accounts using similar language do not prove one operator.

Friends agree. Communities develop shared vocabulary. People copy jokes. Political factions repeat arguments. Fans swarm the same topics.

Stronger attribution can come from platform account records, shared credentials or infrastructure, moderation findings, distinctive operational mistakes, admissions, payment records, or documented coordination linking the accounts.

Public researchers often lack access to the strongest platform-side evidence, so uncertainty should remain visible.

The core harm is simple.

A sockpuppet does not merely add another comment.

It counterfeits another person.

The operator is no longer saying I agree with myself.

The page is saying other people agree too.

Posted on by

Human audits of randomly sampled public discussions

If you go looking only for creepy bot-like conversations, you will find a creepy bot-like internet.

That is not a measurement.

One way to test claims about synthetic conversation is much less dramatic: choose discussions according to a sampling rule decided in advance, then have human reviewers inspect them without selecting only the suspicious ones.

The boring part is the useful part.

Start with a sample that did not already know the answer

A reasonable audit might define a platform, date range, language, discussion type, and method for randomly selecting threads or comments. The sampling process should include quiet, ordinary, messy conversations as well as obvious spam.

Otherwise the researcher is measuring the contents of a folder labeled “weird stuff I noticed,” not the population of the platform.

Human reviewers can then classify observable characteristics: obvious commercial spam, disclosed automation, copied text, coherent human conversation, unknown or ambiguous authorship, and so on.

The word unknown is important.

A reviewer cannot reliably prove that a polished comment came from a human simply because it sounds natural. Nor can repetitive language alone prove automation.

Research on human recognition of social bots illustrates the problem. A 2024 experimental study asked people to identify bots on the VKontakte social network and found that human labeling itself can be difficult enough to undermine the idea of perfect “ground truth.” See Experimental Evaluation: Can Humans Recognise Social Media Bots?.

Disagreement is data

Suppose three reviewers inspect the same account. One calls it automated, one calls it human, and one marks it uncertain.

Throwing away the disagreement would make the final number look cleaner while hiding the most important fact: the evidence was ambiguous.

A useful audit should report how reviewers were instructed, whether they worked independently, how often they agreed, which categories caused disagreement, and how uncertain cases affected the final estimate.

Researchers can use statistical measures of inter-rater agreement, but the plain-language interpretation matters too. “Reviewers agreed on 92 percent of cases” tells a different story from “half the accounts could not be classified confidently.”

A sample answers a bounded question

Even a careful human audit does not establish how much of “the internet” is synthetic.

It can estimate what appeared in a defined sample from a defined platform under defined conditions. Different languages, communities, dates, recommendation systems, and access states may produce different populations.

That limitation is not a weakness. It is what makes the claim testable.

Dead Internet Theory becomes harder to evaluate when every strange screenshot is treated as representative and every normal conversation is dismissed as an exception.

Random sampling reverses that habit.

Do not ask the internet to show you something spooky.

Ask a sample what is actually there, and leave room for the honest answer: sometimes we cannot tell.

Posted on by

Generated comments used to seed an otherwise empty discussion

An empty comment section tells a visitor something.

Maybe nobody cared. Maybe nobody has arrived yet. Maybe the community is small. Whatever the explanation, zero replies are honest information about the current state of the conversation.

That makes emptiness tempting to fix.

A platform can generate a starter comment, a suggested question, or an apparent reaction so the next visitor does not feel like the first person walking into an empty room. Used transparently, that can be a design tool. Used without disclosure, it becomes simulated participation.

A 2026 Scientific Reports experiment on generative AI in social-media discussions tested interventions including AI-written conversation starters, comment assistance, feedback, and reply suggestions. Some tools increased activity, but the researchers also found tradeoffs involving perceived quality and authenticity. See The impact of generative AI on social media: an experimental study.

The important word there is experimental. Participants were studying AI-assisted discussion, not being tricked into believing synthetic participants were ordinary members of a live community.

Seeding changes what the room appears to contain

A genuine conversation starter can be simple: the platform itself posts a clearly labeled prompt asking visitors what they think.

A deceptive version looks different. Several apparently ordinary users arrive first. One asks a question. Another agrees. A third adds a mild disagreement. The exchange exists primarily to make later visitors believe people are already present and engaged.

That appearance matters because human beings use visible participation as social evidence. A thread with comments feels safer to enter than one with none. A product with discussion looks more noticed. A new community with chatter looks more established.

Generated comments can therefore manufacture not just text but social proof.

Disclosure changes the meaning

There is nothing inherently wrong with a machine starting a conversation.

A clearly labeled bot can welcome new users, post daily questions, summarize prior discussion, or keep a support forum organized. Nobody needs to mistake it for a stranger who independently wandered in and cared enough to comment.

The problem is the false inference created when synthetic comments are dressed as ordinary participation.

Ten generated comments do not equal ten interested people. They may represent one operator trying to overcome the cold-start problem of an empty community.

That distinction fits Dead Internet Theory almost perfectly. A page can look socially occupied while containing very little spontaneous human activity.

The useful question is not merely, “Was this comment generated?”

It is, “What does this comment ask me to believe about who is actually here?”

Posted on by

Bot traffic versus bot-authored public conversation

A statistic such as “53 percent of web traffic is automated” sounds like it should tell us how much of the internet is made by bots.

It does not.

It tells us something important, but narrower: machines are responsible for a large share of requests reaching websites and applications. That is a measurement of traffic, not a census of who wrote the visible conversation.

Imperva’s 2026 Bad Bot Report says automated systems generated more than 53 percent of observed web traffic in 2025. That is a remarkable number. It is also easy to misuse.

A request is not a comment

Web traffic measurements usually count HTTP requests. A crawler fetching an article creates a request. A monitoring service checking whether a page is alive creates a request. A scraper downloading prices creates requests. An API client retrieving data creates requests. An attacker testing credentials creates many requests very quickly.

None of those activities necessarily writes anything that another person will read.

Google describes Googlebot as the crawler used by Google Search. It visits pages automatically so they can be indexed. Those visits are bot traffic in the literal sense, but Googlebot is not sitting in the comments pretending to be your uncle.

The mismatch also works in the other direction. One automated account can publish thousands of posts while producing only a modest fraction of a platform’s total network traffic. Meanwhile a single human opening a modern page can trigger requests for HTML, images, scripts, fonts, analytics, advertisements, APIs, and background updates.

The units simply do not map cleanly.

Measuring synthetic conversation requires conversation data

If the question is “What share of public discussion is machine-authored?” the sample has to contain public discussion: posts, replies, comments, messages, or other defined conversational units.

Researchers then need a defensible method for deciding which items are automated. That may involve account behavior, posting tools, content provenance, network patterns, manual review, disclosures, or known ground-truth accounts. Each method has uncertainty and false positives, but at least it measures the thing being claimed.

A traffic report cannot substitute for that work.

This distinction matters for Dead Internet Theory because traffic statistics are often used as if they prove a stronger proposition: if most requests are automated, then most apparent human activity must also be automated. That conclusion does not follow.

The automated web is already enormous. Search crawlers, security scanners, monitoring systems, AI agents, scrapers, spam tools, and malicious bots really do talk to servers all day without a person clicking anything.

That fact is worth studying on its own.

But a machine requesting a webpage and a machine impersonating a person in a conversation are two different events. Counting the first cannot tell us how common the second is.

Posted on by

Hubski: a small discussion community as a counterexample to mass-platform logic

Hubski is useful precisely because it never became enormous.

Mark Katakowski started the site in 2010 while teaching himself programming. By 2013, cofounder Steven Clausnitzer was describing it as an attempt to create a safer place for thoughtful conversation, while Katakowski called the design a mixture of Reddit and Twitter: topics mattered, but a user’s feed was shaped heavily by the people they chose to follow.

That early description survives in the Ann Arbor Observer’s 2013 profile of Hubski. At the time, the founders were talking about thousands of visits rather than millions of users and saying growth came largely by word of mouth.

For a social platform, that sounds almost comically modest.

That is what makes it interesting.

A feed built around people rather than one giant crowd

Hubski did not require everyone to gather around the same front page.

Following a person changed what appeared in your feed. Tags provided another way to find subjects, and comments and shares carried conversations outward. The result could still produce disagreement and cliques — small communities are made of humans, unfortunately — but the architecture did not need every argument to become a site-wide spectacle.

Scale also changed the moderation problem.

A service with a manageable population can depend more heavily on recognizable participants, social memory, and direct intervention. That is not proof that small communities are automatically civil. It simply means they face a different problem than a network trying to moderate hundreds of millions of strangers.

Hubski’s privacy and terms page still reflects some of that deliberately lightweight posture. It says the service does not log users’ IP addresses and does not share user information with third parties except when legally required.

Smallness became a gate

The strongest evidence that Hubski is not chasing mass-platform logic is visible now.

As of 2026 the site is still online, but its signup page says Hubski may be joined by invitation only.

That is nearly the opposite of the normal growth funnel. A conventional platform wants fewer obstacles between a visitor and a new account. Hubski is willing to put an obstacle there.

An invitation does not guarantee a brilliant discussion. It does, however, change the incentives. Growth is no longer the unquestioned objective.

A counterexample, not a miracle

Hubski should not be romanticized.

A small service can lose activity, money, maintainers, or relevance just as easily as a giant one. Its survival does not prove that tiny networks are economically superior, and public pages alone do not tell us how active every corner of the community remains.

What it does prove is narrower and more useful.

The web still contains communities whose success is not measured primarily by becoming universal infrastructure.

Hubski began as a small experiment in thoughtful discussion and, more than fifteen years later, still exists as a recognizably small discussion service. It has not won the internet. It has not needed to.

For a study of supposedly dead online communities, that is worth noticing.

Posted on by

Gather: the trajectory of a discussion network built around adult conversation

Gather launched in November 2005 with a demographic pitch: a social network for adults. Tom Gerace, a Boston entrepreneur who had already founded the affiliate-marketing company Be Free, watched the breakaway services of the day chase teenagers. Gather would court the audience public radio had grown: older, educated, interested in news, books, and culture, and willing to trade arguments rather than party photographs.

The ambition drew media money and partnerships early. American Public Media became an investor and partner, and later described Gather as a social-media service aimed at adults, where members connected around politics, books, cooking, travel, and other interests. An American Public Media description from 2007 said the site had more than 240,000 members and over a million monthly visitors. It was a market the youth-oriented networks were not built around.

Conversation first

Gather presented itself as discussion ahead of networking. Members published essays and short posts, rated one another’s work, and commented across topics from politics to cooking. Gerace’s line was that Gather would do for user-driven media what eBay did for user-driven retail: give a writer a built-in audience instead of a lonely corner of the blogosphere. Trade press described it as “MySpace for a literate audience.” A 2006 MediaPost profile also documented how heavily Gather’s advertising strategy relied on members’ reading, writing, rating, and other behavior. The conversation was community activity and advertising signal at the same time.

Paid to participate

Participation carried a price. Members earned Gather Points for posting and interacting, redeemable for gift cards, and the top writers could earn cash. From 2010, the Gather News Channel paid selected members to write short articles, with extra money tied to page views. It worked well enough: by early 2009 Gather reported roughly half a million members, around 90 percent of them over 25, and more than a million unique monthly visitors. Starbucks chose Gather as an advertising partner specifically for that adult demographic.

Investors matched the enthusiasm. Jim Manzi, the Lotus founder, backed the company early; Allen & Co. joined a January 2006 round; and that November Hearst, McGraw-Hill, and Pilot House Ventures put in $10 million, calling Gather “the leader in social media for grown-ups.”

Sold and emptied

The ownership trail was messier than a single sale. SEC filings show that Health Guru Media bought Gather.com from Skyword in September 2012. Kitara Media later acquired Health Guru Media in December 2013. In Kitara’s accounts, Gather appeared as a domain asset with revenue-linked royalty obligations to Skyword — a strikingly different description from the community of writers who had built the site.

The original social network eventually went offline. Members lost a common home for their writing, comments, and connections; some found one another again through reunion groups and other social networks. What had begun as an adult discussion community ended its corporate life as one asset inside a chain of advertising and media businesses.

What the trajectory leaves behind

Gather’s run shows how an adult-conversation network can come apart. The service genuinely rewarded contribution and demonstrated that an older audience would write, rate, and discuss for recognition and small rewards. But the corporate records eventually describe Gather in the language of domains, revenue and acquisition accounting rather than writers and discussion. A contributor economy can keep conversation alive while it runs; it does not guarantee that the social record will remain central after ownership changes. What remains of Gather is scattered among old captures, rescued essays, and former members who carried the relationships elsewhere.

Posted on by

Slashdot: the changing role of a once-central technology discussion hub

Slashdot went online in 1997 as Rob Malda’s technology-news side project, carrying the slogan “News for nerds, stuff that matters.” Slashdot’s own 20th-anniversary history traces how that small site grew into much more than a news page. For a decade it was the closest thing the internet had to a town square for engineers, sysadmins, and open-source loyalists — the place a technology story went to be argued about in public. The readership treated every technology claim as a leaky sink to be torn apart in full view, which is why a front-page story rarely went unexamined.

A front page run by its readers

What made the site central was mechanics, not charisma. Anyone could submit a story, while Slashdot’s editors selected what reached the front page and wrote the introductions. The crowd’s control operated most visibly in the comments. Slashdot’s moderation system periodically gives selected users points they can spend raising or lowering comment scores, with labels such as “insightful,” “interesting,” “troll,” and “flamebait.” Karma is one factor in who receives moderation access, not a simple rank that permanently grants authority. Historically, meta-moderation added another layer of review over moderation decisions. The system worked well enough that being “slashdotted” became a technical phenomenon of its own, a site that got linked from Slashdot’s front page suddenly collapsing under the traffic. Readers shaped what appeared, ranked it, and argued about it, all on one page.

Reduced centrality, not a shutdown

That position did not survive competing for it. Reddit arrived in 2005 with a simpler vote-up front page, Hacker News in 2007 for the same audience with stricter norms, and social feeds scattered the rest of the conversation onto platforms where nothing was ranked by named votes. Slashdot’s own lifecycle tracked a long series of owners: acquired by VA Linux in 1999, folded into the Geeknet renames, sold with SourceForge to career-site owner Dice Holdings in 2012 for $20 million, and sold again to BIZX in 2016. Comment volume and notoriety peaked years ago. Ownership changes are the easy story to tell, but participation patterns changed the site’s actual role. The anonymous crowd no longer sets the agenda the way it once did, editors assemble most front-page posts from press coverage, and the moderation system that defined the site now serves a much smaller readership.

Still, Slashdot is plainly not abandoned. Slashdot’s current front page continues to turn over with new stories, submissions and scored comment threads. Reduced centrality is not the same thing as death: the community is smaller, older, and no longer the gatekeeper of technology news, but it never stopped doing what it did. The way to tell a fading site from a ghost town is to check whether the residents are still arguing.