Using Dechefr for threat assessment of written communication
Dechefr is designed to support professionals engaged in behavioral threat assessment. It is specifically developed for law enforcement officers, public safety officials, corporate and private security analysts, intelligence professionals, HR specialists, psychologists, psychiatrists, social workers, and academic researchers. Its core function is to facilitate the structured analysis of written communication for indicators of potential violence.
As a digital assistant, Dechefr can be used to analyze a wide range of written material—including social media exchanges, manifestos, weblogs, and other text-based content. Its central concept is to provide a standardized and objective threat assessment, ensuring that all relevant indicators are systematically evaluated. Supported file formats include .txt, .pdf, and .docx.
-
i.Upload or Paste the Target TextBegin by uploading a file or pasting the text you wish to assess.
-
ii.Review the Indicators of Potential Violent BehaviourFor each indicator, click Assess to verify the results. Dechefr will provide an assessment based on automated analysis.
-
iii.VerifyManually inspect the results for each indicator provided by Dechefr. Confirm whether the assigned value is accurate for the context. If necessary, adjust the score.Note: You can check and store supporting sentences/statements or words for each indicator, with the exception of Linguistic Alignment.
-
iv.Save and Manage CasesCompleted assessments can be saved in the Dashboard. You may also retrieve and review previously stored cases for comparison or follow-up evaluation.
-
v.Finalize AssessmentOnce all indicators have been assigned a value, the assessment is summarized in a report. The first section of the report provides an overview of the analyzed text. You can choose to highlight entities such as persons, organizations, and places from the Text Characteristics tab; these will be included in the report. Any sentences, statements, or words you have saved for the various indicators will also be displayed. At the end of the report, you have the option to add your own comments and reflections about the case.
Dechefr automatically identifies and scores various psychological and behavioral indicators based on the content of a written text. While these indicators are machine-generated, it is essential that analysts manually review and confirm their relevance and accuracy within the specific case context. Below is an overview of how each indicator functions in Dechefr, and what role the analyst plays in validating them.
Emotions
Emotions are critical in predicting behavior, as they influence both cognition and action. They also serve as indicators of how an individual interprets and responds to their surroundings. If expressions of anger are prominent in a text, Dechefr will automatically check if anger is present . The analyst must review the relevant sentences to ensure that the emotional tone is correctly identified.
Grievance
Grievance typically stems from a perceived injustice or unfair treatment. This emotional state can fuel a desire—or perceived mission—to correct the wrong and reclaim a sense of justice or status. Research indicates that many lone-actor terrorists and mass murderers are best understood as perpetrators of grievance-fueled violence. Grievances may be personal, ideological, political, or group-based. If such expressions are present, Dechefr will automatically check if grievance is present.
Othering
The repeated use of third-person plural pronouns (e.g., they, them) in reference to a distinct group may signal a strong sense of us-versus-them identity. If the text portrays an outgroup in clearly negative terms, this suggests the presence of othering. Dechefr will detect sings of othering automatically, but the analyst should inspect the context and adjust if necessary.
Leakage
Leakage refers to the communication of intent to harm a specific target. Dechefr leakage automatically and list statements that may indicate leakage. The analyst should read them carefully to determine if they genuinely indicate intent.
Military terminology
Frequent use of military or combat-related language may suggest the individual sees themselves as a warrior, soldier, or chosen agent. This aligns with the warning behavior known as identification. If such terminology is prevalent, Dechefr detects it. The analyst should then verify that the terminology is relevant to the case and reflects this identification mindset.
Influence
The warning behavior identification can also be expressed as identification with a role model. Research has shown that individuals planning to commit targeted violence in many cases have been inspired by previous perpetrators and, in many cases, also mention previous perpetrators in their writings. Dechefr looks for mentions of influential previous offenders and assigns a value automatically. The assigned value should be confirmed manually (and changed if necessary) by the analyst.
Preoccupation
To observe preoccupation, Dechefr provides the 10 most frequently used words and the 10 most frequent collocations phrases. A collocation is two words that often go together. Some collocation examples are "pay attention", "fast food", "make an effort", and "powerful engine". The most frequently used words and collocations may give insights into if the writer is fixated/preoccupied with a person or a cause. If preoccupation is present in the text, Dechefr will produce a value automatically. However, it is important that the analyst confirm.
Linguistic alignment
The linguistic alignment score provides information on what kind of ideology or subculture the writer of the text could be influenced by. Currently there are five different ideologies/ subcultures included: Islamic state, Incel (misogynist online subculture) White supremacy (a belief that white people are superior), counterjihad (beliefs that sees Islam as an existential threat to Western civilization), and, alt-right (partly online movement promoting far-right, white nationalist ideas). Dechefr will produce a value for linguistic alignment automatically that can be changed by the analyst.