About Decipon
Understanding how influence tactics are detected and scored.
Our Philosophy
The Problem with Fact-Checking
Fact-checking is broken. It's slow, politically charged, and creates dependency instead of discernment. The answer isn't telling people what's true—it's teaching them to see manipulation for themselves.
For this to work, three things must be true:
Emotional hooks, coordinated messaging, tribal framing—these aren't subjective. They're patterns that can be detected and scored.
People don't need to be told what to believe. They need to recognize when they're being played. Once you see the technique, you can't unsee it.
Any centralized "truth arbiter" becomes a target for capture. The methodology must be transparent, auditable, and decentralized.
Our Approach
Decipon's Influence Tactics Scoring is a 20-category system that shows you both sides—why content looks manipulative AND why it might be legitimate—so you decide, not us.
What Decipon Does
Decipon analyzes content to surface influence tactics and verification gaps. It identifies persuasion patterns—fear appeals, false dilemmas, urgency triggers, tribal framing—so you can recognize how content attempts to shape your thinking.
What it doesn't do: Decipon does not declare content "true" or "false." It is a tactics lens, not a truth oracle. A high score indicates the presence of influence techniques, not necessarily misinformation. Legitimate persuasion—advertising, opinion editorials, advocacy—may score high without being deceptive.
Who builds Decipon
Decipon and the Influence Tactics Protocol are built by Daniel Bentes at Synapti AS — the company that operates the service (registered in Norway, organisation number 934 968 514).
He designed the 20-category scoring framework, built the hybrid analysis pipeline that
implements it (deterministic ML models for fifteen categories, a temperature-0 LLM tier
for five, plus rule-based heuristics), and authored the open ITP specification. The
methodology, the model cards, and the specification are all public — see the
Synapti organisation on GitHub
and the model repositories under synapti/ on Hugging Face.
Reach the team at hello@decipon.com. Decipon is an independent project; it is not affiliated with any newsroom, platform, or political organisation.
The Scoring Methodology
Decipon's Influence Tactics Score analyzes content across 20 distinct categories of influence tactics. These categories are grouped into 5 composite factors, each measuring a different dimension of persuasion.
1. Emotional Manipulation
Measures how content leverages emotions to bypass rational evaluation.
- Emotional Triggers: Fear, outrage, guilt, or anxiety invoked without proportionate evidence
- Urgent Action Demands: Pressure to act immediately, discouraging reflection
- Novelty Overuse: Claims of "unprecedented" or "never before seen" events
- Emotional Repetition: Same emotional triggers repeated to reinforce response
- Manufactured Outrage: Anger that appears disconnected from underlying facts
2. Suspicious Timing
Examines whether content appears strategically timed for maximum impact.
- Timing Coincidence: Content released to coincide with events, elections, or crises
- Financial/Political Gain: Narratives that benefit specific powerful interests
- Historical Parallels: Patterns resembling known propaganda or influence campaigns
3. Uniform Messaging
Detects coordinated or artificially amplified narratives.
- Phrase Repetition: Identical phrases appearing across multiple sources simultaneously
- Bandwagon Effect: Appeals to conformity—"everyone agrees" or "the consensus is"
- Rapid Behavior Shifts: Sudden, coordinated adoption of symbols, hashtags, or positions
4. Tribal Division
Identifies content designed to create or exploit group conflict.
- Us vs. Them Dynamic: Clear in-group/out-group framing with moral judgment
- Simplistic Narratives: Complex issues reduced to good vs. evil frameworks
- False Dilemmas: Presenting only two extreme options when others exist
5. Missing Information
Evaluates whether content omits context or uses flawed reasoning.
- Context Omission: Alternative perspectives or relevant facts excluded
- Authority Overload: Over-reliance on credentials without substantive evidence
- Suppression of Dissent: Critics dismissed, silenced, or labeled rather than addressed
- Cherry-Picked Data: Selective statistics that misrepresent the full picture
- Logical Fallacies: Arguments containing structural reasoning errors
- Framing Techniques: Language choices that shape perception before evidence is considered
How Scoring Works
Each of the 20 categories is scored from 1 to 5 based on the strength of detected indicators. These scores are aggregated into the 5 composite factors, which are then combined into an overall Influence Tactics Score from 0 to 100.
Important: A high score does not mean content is false. It means the content employs techniques commonly associated with persuasion and influence. Use the score as a prompt to verify claims independently, not as a verdict.
How we validate it
Decipon is a working implementation, not a finished science project. We would rather state the gaps plainly than over-claim accuracy on a tool that scores manipulation. The short version:
- It is consistent with itself. The same content always produces the same scores — the ML tier is deterministic, the heuristics are formula-driven, and the LLM tier runs at temperature 0.
- The ML scorer benchmarks well against an academic corpus. Against the SemEval-2020 Task 11 propaganda-technique dataset it reports a mean absolute error of about 0.368 on a 1–5 scale — for the subset of categories that map onto that taxonomy, not the full 20.
- The score ordering matches editorial intuition. Across a calibration spectrum, a Statistics Norway data release scores near the floor, mainstream reporting low, state media moderate, and partisan opinion high.
- What we have not done yet: no peer-reviewed validation study, no independent third-party benchmark, no inter-annotator agreement study against human expert raters, and no Norwegian-language validation corpus.
- What is planned: a fully designed study — “Empirical Validation of Multi-Dimensional Influence Technique Scoring in Nordic Digital Media” — covering 150–200 Norwegian articles with three expert annotators each and pre-registered agreement thresholds. It is designed but not yet funded; Synapti AS is seeking academic co-investigators.
The full account — the benchmark caveats, the disclosed variance on the LLM-scored categories, the complete not-yet-validated list, and the study design — is on the protocol page. Researchers and institutions interested in collaborating can email hello@decipon.com.
Research foundations
The 20 categories are not invented out of thin air — each operationalizes established work on persuasion, propaganda, and computational manipulation. The core sources:
- Cialdini, R. B. (1984). Influence: The Psychology of Persuasion. William Morrow & Company (revised & expanded eds.: Collins 2006, Harper Business 2021).
- Ellul, J. (1965). Propaganda: The Formation of Men's Attitudes. Alfred A. Knopf (trans. K. Kellen & J. Lerner).
- Jowett, G. S., & O'Donnell, V. (2018). Propaganda & Persuasion (7th ed.). SAGE Publications.
- Bradshaw, S., & Howard, P. N. (Oxford Internet Institute, Computational Propaganda Project) (2019). The Global Disinformation Order: 2019 Global Inventory of Organised Social Media Manipulation. Working Paper 2019.3, Project on Computational Propaganda, Oxford Internet Institute, University of Oxford.
- Da San Martino, G., Barrón-Cedeño, A., Wachsmuth, H., Petrov, R., & Nakov, P. (2020). SemEval-2020 Task 11: Detection of Propaganda Techniques in News Articles. Proceedings of the Fourteenth Workshop on Semantic Evaluation (SemEval-2020), International Committee for Computational Linguistics.
- Pratkanis, A. R., & Aronson, E. (2001). Age of Propaganda: The Everyday Use and Abuse of Persuasion (rev. ed.). W. H. Freeman (revised ed.; 1st ed. 1991).
The full bibliography, with a note on how each source maps onto the protocol's categories, is on the protocol page.
The framework's initial structure was inspired by publicly discussed work by Chase Hughes on influence and behavioral analysis; it has since been grounded in the academic literature above. Decipon is an independent project — the ITP is an independent specification, not affiliated with or endorsed by Chase Hughes or any related organizations.