Predictive Processing
Therapy (a Claude experiment)
Hate Speech in Germany - Examples and how to test
Ever wondered about freedom of speech or "hate speech" in Germany? I accidentally came across a website that lets you assess statements.
Understanding Nulu.io's Hate Speech Assessment Tool
This application demonstrates how www.nulu.io, a German AI company, assesses potential "hate speech" in text. Since hate speech isn't a precise legal concept and definitions vary, this tool provides insight into how such assessments work in practice.
About Nulu
Nulu is a German AI company that primarily serves public sector clients, including schools. Their hate speech detection tool allows users to submit text and receive an automated assessment via email.
How the Tool Works
Users enter their email address and a statement. The system then evaluates the text and sends back a risk percentage indicating the likelihood that the statement could:
Contain hate speech elements under German law
Face legal consequences (police investigation, fines, or imprisonment)
The assessment is based on German legislation covering personal insults and other relevant laws, including special protections for politicians under §188.
Limitations
Nulu.io provides only a single percentage (like 65%) without explaining their reasoning or methodology.
Potential Applications
While the company doesn't specify intended uses, possible applications might include:
Self-monitoring/censoring to avoid legal issues
Reporting concerning content to authorities such as the police
Working with "trusted flagger" organizations
Starting lawsuits
About Trusted Flaggers
A complementary concept is that of "trusted flaggers". They are one way to report content.
The European Union established the "trusted flagger" concept, where designated NGOs and other organizations can report hate speech or so-called "disinformation" to authorities.
At least in Germany, the oddly named trusted flagger "Respect" is a political organisation, founded by the German Greens, and funded by government programs. My personal view is that it is a censorship organisation with a political direction.
This may not be the case for other countries, for example Ireland. Here, it is the genuinely respected Central Bank of Ireland.
Adding Context with AI
Since Nulu.io doesn't provide reasoning for their assessments, this mini application used Claude.ai to suggest possible criteria that might have influenced the risk percentage. Both AI systems likely draw from similar knowledge bases, making this a reasonable approach to understanding the assessment logic.
Note: this is an AI generated application
References / Sources
The website providing the percentage assessment
https://www.nulu.io
About the statements:
Most of them taken from statements reported in the press, that I overheard in conversations or on TV , or having been made known through law suits that became public). Some of them I constructed to test the system with varying formulations.
Except a few of the statements at the top, these are not views I would express in this way, although in see the consequences of uncontrolled immigration from Afghanistan, Syriah etc. Here, I am informed by the official crime statistics, and by personal knowledge of the situation at schools.
What is the system to report "hate speech" in Germany
Despite the fact that "hate speech" is not a legal concept, and only closely defined, there is quite a network of ways to report it to the authorities.
https://www.perplexity.ai/search/in-germany-how-can-i-report-ha-1PmoSn8iRFOXAwC4gtG4Cw