Framed Mind

Spot the frame before it shapes your opinion

Our nudge, “Framed Mind,” makes potential framing effects in AI-generated responses visible. Responses generated by Large Language Models (LLMs) are color-coded depending on whether they are predominantly positive, neutral, or negative in tone. Positive statements are marked in orange, neutral statements in green, and negative statements in blue.
The color-coding serves as a visual guide and immediately draws users’ attention to potential biases. This is intended to prevent AI-generated content from being accepted uncritically. The nudge intervenes directly during the use of AI systems and promotes a more conscious engagement with the information presented.

Accept the cookie setting to see Panopto content.

Unlock Panopto

What does the topic mean?

Generative AI systems such as ChatGPT are increasingly being used in educational settings. Pupils and university students use these applications for research, writing texts, and supporting their learning processes. AI-generated responses are o en perceived as objective
and neutral.
In reality, even small differences in the wording of a question or response can lead to information being presented more positively or negatively. This phenomenon is known as framing. As AI systems have a growing influence on information processing and opinion formation, raising awareness of such biases is becoming increasingly important.

Goal of the nudge

The goal of the nudge is to raise users’ awareness of potential biases in AI-generated responses and encourage a more critical approach to the information provided. Many people trust AI systems and adopt their statements without questioning them sufficiently. By making framing effects visible, users are encouraged to reflect more consciously on content and consider different perspectives more carefully. In the long term, this aims to strengthen media and AI literacy while reducing the risk of uncritical opinion adoption.

    Needs analysis

    • Increasing use of AI in education
      Generative AI is being used more and more frequently by pupils and university students for learning and research purposes.
    • Lack of awareness of biases
      Many users are unaware that AI responses can be influenced by wording and framing.
    • Influence on opinion formation
      AI-generated content can influence attitudes and decisions without users consciously noticing it.
    • Need for media and AI literacy
      Educational institutions are looking for ways to promote a reflective and responsible use of AI.

    Cause analysis

    • Framing Effect

      The way information is presented influences how it is perceived and evaluated.

    • Confirmation Bias

      People tend to prefer information that confirms their existing beliefs.

    • Automation Bias

      Statements generated by AI systems are often perceived as objective and correct and
      are therefore questioned less critically.

    • Cognitive Offloading

      Users delegate thinking and evaluation processes to AI instead of assessing
      information independently.

    • Social Proof

      The widespread use of AI within social environments increases trust in AI-generated
      content.

Target Group

The primary target group of the nudge consists of pupils and university students. They use AI applications particularly frequently in learning contexts and are simultaneously in a phase where opinion formation and information evaluation play an important role. In addition, the nudge may also be beneficial for teachers and other users of generative AI.

Added value of the nudge

  • Makes framing visible

    Users can more easily recognize that AI responses may contain different perspectives and tendencies.

  • Encourages cri cal thinking

    The labeling encourages users to question information more consciously rather than
    accepting it without reflection.

  • Strengthens media and AI literacy

    Users learn to use AI systems in a more reflective and responsible manner.

CONTACT US

Elizaveta Alberg

elizaveta.alberg@stud.hshl.de

Christiane Duong

christiane.duong@stud.hshl.de

Jara Gall

jara.gall@stud.hshl.de

Ronja Hill

ronja.hill@stud.hshl.de