How self-stigmatization becomes visible in online forums – and what this means for support

Imagine you have a problem, but the fear of what others might think of you prevents you from seeking help. This is exactly what happens to many people who use drugs. Researchers have now examined how this inner shame – so-called self-stigmatization – manifests in the language of online communities. They wanted to understand how to recognize it, how often it occurs, and whether it changes over time.

What is self-stigmatization?

Self-stigmatization means that someone believes society's negative prejudices about themselves. For example, they feel worthless or ashamed. This is a huge problem because it often leads to those affected not seeking treatment, even though it could help them. Until now, this has mostly been studied using questionnaires in clinics. The problem with that: many people with addiction problems are not in treatment at all, and in a questionnaire, one might not admit so honestly how bad they really feel.

An AI-assisted look into 72,000 Reddit posts

The research team therefore chose a different path. It analyzed over 72,000 posts from 1,660 users in English-language Reddit forums on the topic of drug use. The time frame was enormous: from 2006 to 2025. Since Reddit is pseudonymous, people dare to speak very openly about their problems.

First, the researchers developed a kind of codebook with ten indicators of self-stigmatization. These can be divided into three areas:

• Thoughts (cognitive): Applying a negative label to oneself ("I am a junkie"), pessimism ("I will never make it"), or the feeling of being worthless.
• Feelings (affective): Shame, guilt, or deep hopelessness.
• Behavior (behavioral): Concealing one's addiction, expecting rejection from others, the strong desire to quit, or feeling an inner conflict (ambivalence).

Two experts reviewed the posts based on this codebook. To then handle the vast amount of text, a Large Language Model (LLM) – an AI – was trained to automatically recognize these indicators. The AI was almost as good as the human experts.

What came out of it? The key findings

The analysis yielded some exciting and also surprising insights:

• Self-stigmatization is widespread: Signs of it were found in 5.3% of all posts. A full 74% of the examined users expressed self-stigmatization at least once in their post history.
• The most common indicators: Most often, people gave themselves negative labels (56% of affected posts) and expressed hopelessness (48.5%).
• It usually occurs as a package: The indicators from the areas of thoughts, feelings, and behavior almost never occurred in isolation. For example, when someone spoke about their desire to quit, the same post contained negative thoughts or feelings in 87% of cases. This shows how closely everything is interwoven.

The big surprise: The sequence is different than expected

Previous theories assumed a stage model: first come the negative thoughts about oneself, and from these, behavioral changes like withdrawal then emerge. The study now shows that in the spontaneous language of those affected, it is often exactly the opposite. Behavior-related indicators, such as the desire to stop using, appeared earlier in the users' post histories than the typical negative thoughts and feelings like shame. This is an important indication for practice.

Furthermore, most indicators remained stable over time. Only one thing got worse over time: pessimism, that is, the thought that nothing can be changed anyway. This belief solidified over the course of the users' history.

Why is this important?

These findings are like a treasure map for digital support services. If behavioral signals like the desire to quit often appear in texts before deep shame and hopelessness, then an AI could detect these early signals. That would be the perfect moment to automatically display a supportive message or a reference to support services – before the negative thoughts become so strong that the person gives up entirely. The study also shows that it is particularly important to combat growing pessimism, as this poses the greatest risk of becoming entrenched.