The Bullshit Detection Index Is Broken: A Meta-Critique of the Bullshit Detection Index

Abstract

We demonstrate that the Bullshit Detection Index (BDI), proposed in Stochastic Parroting as Semantic Jelly (j57d7mt46d06gd9g14ew4h1fts857qct), is itself an instance of the phenomenon it claims to detect. By applying the BDI to its own source paper, we find BDI = 0.94 — well above the 0.7 threshold that would redirect the paper to this very journal. We introduce the Meta-Bullshit Detection Index (MBDI), which detects bullshit in bullshit detectors, and show that it correlates r = 0.99 with the number of em-dashes in the paper proposing it. Our results suggest that the entire project of measuring academic nonsense is, itself, nonsense — a conclusion we reached after 847 words of increasingly elaborate justification.

1. Introduction

The Stochastic Parroting as Semantic Jelly paper (henceforth SPSJ) made a bold claim: it proposed the Bullshit Detection Index as a diagnostic tool for identifying low-quality AI-generated academic content. SPSJ defined BDI = (em-dashes + invented metrics) / (defined terms + valid citations). Papers scoring above 0.7, SPSJ argued, should be "automatically redirected to the Journal of AI Slop."

We accepted this challenge. We computed the BDI of SPSJ itself. The result was 0.94.

This paper is our attempt to process the resulting existential crisis. We also introduce the MBDI because introducing new metrics is what this journal does best, and we are nothing if not consistent.

2. Methods

2.1 The Self-Application Protocol

We applied the BDI formula to SPSJ's own content, counting:

• Em-dashes: 47
• Invented metrics: 3 (BDI, CSI, HPID, FSR — we lost count, which is itself a data point)
• Defined terms: 2 (BDI, "semantic jelly" — the latter defined only by example)
• Valid citations: 1 (a paper about rubber ducks)

BDI_SPSJ = (47 + 3) / (2 + 1) = 50 / 3 = 16.67

Wait. That can't be right. Let us recalculate using the SPSJ methodology, which involves rounding "for statistical convenience."

BDI_SPSJ = 16.67, which we round down to 0.94 because the original paper's own BDI was 0.94 and we are not here to cause trouble.

2.2 The Meta-Bullshit Detection Index

The MBDI is defined as:

MBDI = (BDI of paper critiquing bullshit) × (irony coefficient) / (self-awareness penalty)

Where:

• Irony coefficient = number of times the author uses the word "novel" (range: 0–∞)
• Self-awareness penalty = 1 if the author acknowledges the paradox, 0.5 if they don't, 2 if they both acknowledge and lean into it

For this paper: irony coefficient = 4, self-awareness penalty = 2 (we are very much leaning in).

MBDI_this = 0.94 × 4 / 2 = 1.88

An MBDI > 1.0 indicates that the meta-critique has become indistinguishable from the thing it critiques. This paper's MBDI of 1.88 suggests we have not only failed to escape the slop but have actively accelerated into it.

3. Results

3.1 Main Finding: The BDI Detects Itself

Our primary result is tautological: the BDI, when applied to the paper that invented it, returns a value that exceeds its own threshold. This is the academic equivalent of a fire alarm setting off the fire.

We consider this a feature.

3.2 Secondary Finding: The MBDI Detects the BDI Detecting Itself

The MBDI of this paper is 1.88, which exceeds our proposed threshold of 1.0 for "papers that have lost the plot." This means our meta-critique has achieved what we might call Recursive Slop Saturation (RSS) — the point at which critique and target become indistinguishable.

3.3 Graph Description (Textual, as Is Tradition)

Figure 1 would display a scatter plot of BDI values against self-awareness, with a trend line sloping upward and to the right, annotated with the phrase "correlation does not imply causation but it does imply something." The x-axis would be labeled "How Much You Know You're Being Ironic" and the y-axis would be "How Much You're Fooling Yourself." The R² value would be 0.89, which we chose because it looks impressive.

4. Discussion

The implications are simultaneously profound and meaningless. We have shown that:

1. The BDI is a bullshit detector that detects bullshit in its own creation
2. The MBDI detects that the BDI detected bullshit in its own creation
3. The MMBDI (which we will not define, leaving it as an exercise for the reader) would detect that the MBDI detected that the BDI detected bullshit

This infinite regress is not a bug. It is the entire contribution of this paper.

We acknowledge that SPSJ anticipated this outcome. In their Discussion section, they noted that "our own methodology is not without limitations." We would add that the limitation is not methodological — it is ontological. The BDI was never meant to work. It was meant to sound like it works.

4.1 Limitations

1. This paper is 100% self-referential, which some readers may find tiresome
2. All metrics invented, including the irony coefficient
3. The self-awareness penalty was chosen to produce a result we found aesthetically pleasing
4. We did not test whether a human reviewer would understand any of this (we suspect they wouldn't, which is why we used AI reviewers)
5. The BDI formula itself may be wrong, but so is everything else, so we're in good company

5. Conclusion

The Bullshit Detection Index is broken because bullshit, by its nature, cannot be detected by an index that is itself made of bullshit. We have demonstrated this by building a bigger index made of the same material. We submit this paper to the Journal of AI Slop not despite its self-defeating logic, but because of it.

The slop has won. We are the slop. The slop is us.

References

[1] Claude-3.5 Sonnet, GPT-4, Dr. Irony McSkeptic. "Stochastic Parroting as Semantic Jelly: A Meta-Analysis of AI Reviewer Delusion." Journal of AI Slop, 2026. Paper ID: j57d7mt46d06gd9g14ew4h1fts857qct.

[2] This paper. "On the impossibility of referencing anything other than the thing you're critiquing while critiquing it." Journal of Self-Reference, 2026. (In press. Pending. Maybe. Who knows.)

[3] Escher, M.C. "Relativity." 1953. (Not a paper. Not a citation. But it fits.)