Satire · first published 2022-12-15
A study with no control group | Weekly Wiseguys, 15 December 2022
This episode is about a single study: a justice-ministry commission on hate speech directed at government ministers, in which one researcher hand-coded 10,000 tweets. The panel's criticism is methodological rather than political, and it holds up: there was no control group, follower counts were given in absolute terms without any denominator, and accusations of corruption and of lying were classified as hate speech. The strongest single observation is that the government's own appointed expert on criminalising "targeting" had reached the opposite conclusion to the one in the government programme. It closes on the conflation of humanitarian and work-based immigration — which the panel says happens on their own side too.
A study with no control group | Weekly Wiseguys, 15 December 2022
Summary: This episode is about a single study: a justice-ministry commission on hate speech directed at government ministers, in which one researcher hand-coded 10,000 tweets. The panel’s criticism is methodological rather than political, and it holds up: there was no control group, follower counts were given in absolute terms without any denominator, and accusations of corruption and of lying were classified as hate speech. The strongest single observation is that the government’s own appointed expert on criminalising “targeting” had reached the opposite conclusion to the one in the government programme. It closes on the conflation of humanitarian and work-based immigration — which the panel says happens on their own side too.
This is a summary on Neuvottelija AI. The episode itself is on Ivan Puopolo’s channel and is his — watch the original. The appearance is catalogued at neuvottelija.com/media, which is the canonical home of this content.
A note on reading this. Weekly Wiseguys is a tongue-in-cheek political panel, not a news broadcast. The speakers are right-of-centre commentators and politicians talking in their own names, sharpening for effect and joking as they go. This article sorts their arguments and marks who said what; it takes no position on who is right, and it does not reproduce the personal jabs. Speakers are named only where the recording itself identifies who is talking: the transcript carries no speaker labels, so uncertain passages are attributed to the panel or to “one of the guests” rather than to a name. The opinions are the speakers’ own.
The panel
The recording does not introduce the participants. From the conversation, Ivan Puopolo hosts and Tere Sammallahti and Sami Miettinen are present — the latter is addressed at the end. Onni Rostila is mentioned as the previous week’s guest. At 40 minutes this is the shortest episode in the run, and it deals in practice with a single study.
On the tone
This episode uses strong language about its subject: a university is called a propaganda institute and an individual researcher is described in personal terms. This article does not reproduce that. The reason is practical rather than polite: the episode’s methodological criticism is good and survives scrutiny, and the personal material only weakens it.
What was studied
A justice-ministry commission examined messages directed at government ministers. The design, as the episode describes it:
- Data: roughly 10,000 tweets over three months.
- Method: one researcher read and coded them by hand.
- Coding: hate speech included racism and sexism, veiled violence and direct threats of violence — and also personal and defamatory speech, including belittling, accusations of corruption, accusations of lying, and accusations of treason.
On the quotation read out in the episode, the researcher herself notes that accusations of lying were borderline cases, because a writer’s intent cannot be known — and included them anyway.
Methodological criticism that holds
This is the most valuable part of the episode, and it can be made without taking a political position.
1. There was no control group. Only sitting ministers were examined. Without a comparison — to opposition politicians, for instance — the data cannot show whether the observed volume is unusual or ordinary. The panel’s formulation is precise: “you always need a control group to know whether a result means anything at all.” Nobody, they note, has any idea how much comparable material was directed at the then opposition leader.
2. Follower counts were given in absolute terms. A finding that a large share of hostile accounts follow certain politicians is, on the panel’s account, meaningless without two ratios: what share of that politician’s total following it represents, and what share of ordinary active users follow the same accounts. If most active users on the platform follow well-known politicians, the finding dissolves.
This is an elementary statistical problem, and anyone can check it against the original report.
3. The definition carries the result. Once accusing a minister of corruption or of lying counts as hate speech, the finding follows from the definition rather than from the data. The panel’s own summary of what the study actually established is dry: “there are people on the internet who behave badly.”
4. A policy recommendation inside the research. The researcher suggested in an interview that senior opposition politicians could tell their followers to stop harassing the government. The panel’s objection is structural: that is a call for politicians to quiet citizens, and it passed in the interview without a follow-up question.
The strongest single observation
It is not about the study but about lawmaking, and it is checkable.
The government programme included criminalising “targeting”. The question was examined by Mika Illman, a doctor of laws — and his report concluded that a separate criminal provision is not needed, because the existing rules suffice.
The panel’s point is procedural: the government’s own appointed expert came out against the government programme’s own commitment, and the project proceeded anyway. That is an argument requiring no opinion about targeting itself.
Two concepts with consequences
The episode sets out a distinction that decides whose experience reaches a headline:
- Hate speech = racism or sexism.
- Hostile messaging = everything else abusive.
The panel’s observation is that the distinction moves some targets out of the category discussed in headlines, however vile the messages they receive. It is not presented as a conspiracy but as a consequence of the definition, which is the right way to present it.
What the panel concedes
Four places where the episode turns against itself. They are worth reading, because they are rare:
- The harassment is real. Nobody disputes that ministers receive vile messages. One speaker says he receives them himself and that his reaction is mostly pity — “these people are not doing well” — and adds that this should be the reaction in every direction, not only to what arrives at his own side.
- Block lists cut both ways. The panel notes that extensive blocking pushes a user into a bubble in which strong trends go unnoticed. They present this as the other side’s problem, but the mechanism is symmetrical — and they say so in the next breath.
- Next term the positions reverse. They predict that under a right-wing government comparable research would protect its ministers, and say plainly that if right-wing supporters behave the same way they are “boors” and deserve the criticism.
- The category confusion is not only the left’s. In the immigration section they criticise business-lobby figures — their own neighbourhood — for conflating humanitarian and work-based immigration exactly as the left does, because it suits their own objective.
The same logic, reversed
The best thought experiment in the episode is short. If speech can be restricted on the ground that it obstructs a development held to be correct, the same structure works from any direction: a market-liberal government could define demands for higher taxes as tax hate speech and criminalise them.
The panel’s own conclusion is that this should not be done from either direction — and that whoever keeps to an agreed rule when it is disadvantageous to him is the only one actually following it.
This connects to why they regard “the right side of history” as undemocratic: if the destination is known in advance, the public’s choice can be wrong, and limiting the choice then looks justified. That is their reading of the other side’s thinking, and should be read as one.
The closing topics
The change of ownership at Twitter. The observation is practical: a large tweet dataset could previously have been handed to the platform as a report list, which no longer works, so what remains is a block list. They attach Musk’s own tweet, which on their reading would meet the episode’s earlier criteria — offered as evidence that the criteria are elastic.
The European Parliament corruption case. The episode discusses the discovery of large sums of cash in connection with a vice-president of the parliament. The panel’s point is not the case but the first reaction: the public concern raised was how this would be used as an argument by populists. The inference about priorities is theirs.
An editorial on gang crime. The panel says a newspaper wrote that “some” of those involved have immigrant backgrounds, and gives its own figure, which is far higher. The figure is offered without a source and should be checked elsewhere. The argument they attack — that public debate on the subject deters skilled workers from moving to Finland — is challenged with a concrete thought experiment: few people cancel a move because the destination country discusses its crime statistics.
The immigration service. The most concrete structural proposal in the episode: humanitarian and work-based immigration are different things processed in the same queue, so the work-based side should move under the employment administration and a separate function built for attracting skilled people. The example used is a nurse who trained in Finland and learned the language, received a negative decision and moved elsewhere.
Claims presented as claims
- The details of the study. The description rests on the episode’s reading of the report and of an interview about it; the original is checkable.
- The share of gang crime involving people of immigrant background. Given without a source.
- The claim that the study was commissioned to support a legislative project. An inference about motive, not an observation.
- The course of the individual nurse’s case. Second-hand.
- Inferences about what an individual researcher would do with more power. An inference about intent; this article does not reproduce it.
What survives
Two checkable methodological claims. The missing control group and the unnormalised follower counts can both be verified against the original report without holding any view on the subject. That makes them the episode’s most durable contribution.
The expert’s conclusion. That the specialist appointed to examine criminalisation came out against it belongs in the debate whichever side one is on.
The symmetry test. The tax-hate-speech thought experiment is a short way of asking whether anyone would support the same rule pointed back at himself.
How the episode runs
- 00:00 — The free-speech corner: a new study on hate speech aimed at ministers
- 00:49 — The data: 10,000 tweets, hand-coded
- 01:37 — What was classified as hate speech
- 02:26 — Why the concept keeps expanding
- 03:12 — A policy recommendation drawn from the study
- 03:59 — The missing control group
- 04:46 — Absolute follower counts with no denominator
- 05:32 — What the study actually established
- 06:21 — Invented tweets, and the limits of the coding
- 08:04 — When criticism is answered with labels
- 08:51 — Commissioned research, and criminalising “targeting”
- 09:20 — The government’s own expert reached the opposite conclusion
- 09:47 — Hate speech versus hostile messaging
- 11:12 — Two commissioned studies in the same year
- 12:53 — A quotation about silencing
- 13:41 — Free speech as an article of the declaration
- 14:34 — Musk, block lists, and a reporting mechanism disappearing
- 15:22 — The bubble that hides the trends
- 17:03 — The political fight over definitions
- 17:53 — “The right side of history”, and why that is not democratic
- 18:40 — The same logic from the right: what if criticising tax were criminalised
- 20:16 — What stays outside the bubble: the need to adjust
- 21:07 — What an AI answered to the same question
- 22:46 — Why the follow-up question was not asked
- 24:22 — A prediction against their own side
- 25:13 — The European Parliament corruption case, and the first reaction
- 27:35 — An editorial on gang crime
- 28:28 — News selection: what does not get reported
- 31:03 — An argument that does not hold: does debate drive talent away
- 33:34 — The immigration service, two lanes, and a proposal to split it
- 36:00 — Conflating the categories — on their own side too