Location: Lecture hall FAL 158 (Bürohaus am Falkenbrunnen, Chemnitzer Strasse 46 b, Dresden). Everybody is welcome.

As alternative platforms and digital sovereignty for Europe are separately needed, we aim to question the current status quo and explore how this technology offers untapped opportunities for an improved information landscape and participatory democracy lived online.

A number of political behaviours appear to be influenced by the use of digital media, such as increasing polarization or declining trust in institutions.

Increasing interconnectedness has led to more self-organized public debates, platforms and their algorithms have gained new power over discourse, and generative AI has made content fabrication easier than ever.

Abstract: Information and communication technology has undergone dramatic developments over the last two decades.

You have been cancelled. So

The talk is associated with our Master's program "Human Performance in Socio-Technical Systems (HPSTS)", but open to the public.

Language: German

Venue: Bürohaus am Falkenbrunnen (Chemnitzer Straße 46b), FAL 158

Time: Dec 12th, 1:00 pm

The next important step on this journey will be to develop clearer theoretical ideas of the individual descriptor properties, and of the ways in which they are functionally related to one another. Exciting, but challenging. We're working on it...

Despite their (current) shortcomings, LLM-based ratings of descriptor properties may soon help us overcome this grave limitation because *any set of items* (e.g., from an already published study) may be easily assessed this way, a posteriori.

However, to this day, most psychometric research still ignores them, making data analyses a lot less informative than they could be.

Descriptor properties have been the subject of psychological research for decades, including seminal contributions by (e.g.) Edwards (1953), Funder & Dobroth (1987), and John & Robins (1993).

Going beyond previous work, we also used LLM-based ratings of the same properties, finding that they resemble human-based ratings quite a bit, but should not (yet) be treated as being entirely interchangeable with them.

Using a new sample of participants who generated a new sample of terms, most of the previous findings that we targeted were indeed replicated (e.g., the pattern of correlations among the six descriptor properties).

Investigating these, and the ways in which they interact with other factors (e.g., in the perceiver, the target, or in the relationship between perceiver and target), is crucial because most psychometric assessment (e.g., by means of questionnaires or interviews) uses natural-language items.

We looked into a variety of properties that distinguish person-descriptive terms in the natural language ("person descriptors") from each other: Social Desirability, Observability, Abstractness, Stability, Importance, and Base Rate.