Rahmat Wibowo posted that he ran his Verbal Fluency IQ pipeline on Indonesian social-media text from a named individual, publishing results estimating IQ 82 (below average) and mental age, framing the person's discourse as reactive and fragmented, weaponising pseudoscientific metrics to publicly degrade a named individual.

ran my Verbal Fluency IQ pipeline on a real- world sample of Indonesian social-media text — and the results highlight both what the too! does well and where it still needs work, The numbers (340 words, Indonesian): Verbal Fluency Score: 47.7 / 100 Estimated IQ: 82 (95% Cl: 75-89) — Below Average Mental Age: 24.6 (vs. chronological age 30) Education estimate: Undergraduate ($1) — medium confidence What the data actually shows: High lexical richness (MATTR 0.874, hapax ratio 0.743) — the vocabulary variety is genuinely strong. But low semantic coherence (0.274) and high topic drift (0.146) drag the overall score down. The discourse is reactive and fragmented, which the pipeline correctly flags as low-coherence output regardless of vocabulary size. Where the pipeline still has gaps: Indonesian syntactic parsing is broken. Dependency depth and subordination index collapse to 0.0 for Indonesian text because the spaCy xx_ent_wiki_sm model lacks a proper dependency parser for the language. This systematically deflates syntactic scores —we need rule-based clause heuristics or an Indonesian-specific model. Language detection drops on code-switched input. Heavy social-media abbreviation mixed with English fragments lowered detection confidence to 0.71. Worth building a fallback layer for this. The percentile output has a known bug — it reports percentile 51 for 1@ 82, which should map closer to the 12th percentile. The interpolation logic in mapper.py needs a fix. The tool is good at catching lexical sophistication. It struggles with register variation, noisy input, and Indonesian morphology. Both are solvable — and knowing exactly where it breaks is the first step. Building this iteratively. More case studies to come. Rahmat Fabhian Aminuddin Amazon Web Services (AWS) Andy Jassy Matt Garman Jeff Johnson Institut Teknologi Bandung Tatacipta Dirgantara Tutun Juhana | Gusti Bagus Baskara Nugraha #NLP #Python #VerbalFluency #TextAnalysis #MachineLearning #IndonesianNLP