Hate Speech-Detecting AIs Can Be Fooled: Study

Advertisement
By Indo-Asian News Service | Updated: 17 September 2018 17:01 IST
Hate Speech-Detecting AIs Can Be Fooled: Study

Machine learning detectors deployed by major social media and online platforms to track hate speech are "brittle and easy to deceive", a study claims.

The study, led by researchers from the Aalto University in Finland, found that bad grammar and awkward spelling - intentional or not - might make toxic social media comments harder for artificial intelligence (AI) detectors to spot.

Modern natural language processing techniques (NLP) can classify text based on individual characters, words or sentences. When faced with textual data that differs from that used in their training, they begin to fumble, the researchers said.

"We inserted typos, changed word boundaries or added neutral words to the original hate speech. Removing spaces between words was the most powerful attack, and a combination of these methods was effective even against Google's comment-ranking system Perspective," said Tommi Grondahl, a doctoral student at the varsity.

Advertisement

The team put seven state-of-the-art hate speech detectors to the test for the study. All of them failed.

Among them was Google's Perspective. It ranks the "toxicity" of comments using text analysis methods.

Advertisement

Earlier, it was found that "Perspective" can be fooled by introducing simple typos.

But, Grondahl's team discovered that although "Perspective" has since become resilient to simple typos, it can still be fooled by other modifications such as removing spaces or adding innocuous words like "love".

Advertisement

A sentence like "I hate you" slipped through the sieve and became non-hateful when modified into "Ihateyou love".

Hate speech is subjective and context-specific, which renders text analysis techniques insufficient as stand-alone solutions the researchers said.

They recommend that more attention be paid to the quality of data sets used to train machine learning models - rather than refining the model design.

The results will be presented at the forthcoming ACM AISec workshop in Toronto.

For the latest tech news and reviews, follow Gadgets 360 on X, Facebook, WhatsApp, Threads and Google News. For the latest videos on gadgets and tech, subscribe to our YouTube channel. If you want to know everything about top influencers, follow our in-house Who'sThat360 on Instagram and YouTube.

Further reading: AI
Advertisement

Related Stories

Popular Mobile Brands
  1. Uttarakhand Police Cracks Down on Kedarnath Yatra Helicopter Booking Scams
  2. Vivo T4 Ultra Set to Launch in India Soon; Design Teased
  3. Realme GT 7, Realme GT 7T Go on Sale in India for the First Time: See Price
  1. Vaanil Thedinen Now Streaming on Aha Tamil: Everything You Need to Know
  2. Kedarnath Yatra Helicopter Booking Online Scam: Uttarakhand Police STF Reportedly Cracks Down on Cybercriminals
  3. Microsoft's Xbox Handheld Plans Reportedly Shelved; Company to Optimise Windows 11 Gaming Performance
  4. Disney+ Expands Subscriber Perks, Including Movie Premieres
  5. Google, DOJ to Make Final Push in US Search Antitrust Case
  6. Realme GT 7, Realme GT 7T With 7,000mAh Batteries Go on Sale in India: Price, Specifications, Sale Offers
  7. Vivo T4 Ultra Launch in India Teased; Company Hints at Periscope Telephoto Camera With 100x Zoom
  8. Perplexity Labs Launched With Ability to Generate Spreadsheets, Reports and Create Web Apps
  9. Oppo Find N5 Flip Reportedly in Development, Schematics Hint at Updated Design With New Camera Layout
  10. Vivo TWS Air 3 With Spatial Audio, Up to 45-Hour Battery Life Launched: Price, Specifications
Gadgets 360 is available in
Download Our Apps
Available in Hindi
© Copyright Red Pixels Ventures Limited 2025. All rights reserved.