AI-Assisted Adaptation of the Swedish Version of the Digital Competencies for Applied Psychological Practitioners...

As digital technologies become increasingly integrated into psychological practice, the demand for competencies in digital clinical psychology is growing. Although competency frameworks for digital clinical practice exist, validated instruments to assess these competencies remain scarce. In Sweden, psychology master's students are now being offered digital psychology courses, increasing the need for instruments to measure intended improvements in knowledge and abilities. Using AI to assist with translation procedures can facilitate the adaptation of existing instruments to new national and cultural contexts.

This study aims to test an AI-assisted procedure for the translation and contextual adaptation of the Digital Competencies for Applied Psychological Practitioners (DCAPP) scale into Swedish and to examine the psychometric properties of the translated version in a sample of psychology master's students in Sweden, including a pilot test of the instrument's responsiveness to changes in knowledge and abilities among students attending a digital psychology course.

An AI-assisted adaptation procedure, using ChatGPT and DeepL, was used to translate the DCAPP from English to Swedish. The Swedish version was distributed to psychology master's students during their eighth semester, including those attending an elective digital psychology course. Twenty-four of 55 master's students completed the baseline questionnaire. Of the 14 students enrolled in the digital psychology course, 13 completed the baseline questionnaire and 9 completed the follow-up assessment. Item descriptives, internal consistency, and responsiveness to change were calculated for the scale.

The AI-assisted translation procedure resulted in a translated version of the scale with both high quality and semantic similarity ratings. The Swedish DCAPP demonstrated excellent internal consistency for the total score (α=.96) and also for Knowledge (α=.93) and Abilities (α=.96) subscales. It demonstrated acceptable item distributions, with item-total correlations above 0.30 (range: 0.53-0.87), and mean interitem correlations for the subscales were acceptable but indicated potential item redundancy (Knowledge r=0.48; Abilities r=0.61). Although skewness and kurtosis values were mostly acceptable, pronounced floor effects were observed in both subscales. A statistically significant increase in students' competency ratings was observed at post test (P<.001), suggesting preliminary sensitivity to change.

Using an AI-assisted adaptation procedure to support translation is feasible. The Swedish DCAPP showed promising psychometric properties and preliminary evidence of responsiveness to change. Floor effects may have been due to students' limited digital competencies. Although initial results are promising, further research with larger samples is needed before the psychometric validity of the Swedish DCAPP can be confirmed.