Participants completed 10 sessions of sham HD-tDCS stimulation over 5 consecutive days (two 20-minute sessions per day, with a 3-hour interval between sessions). Stimulation was delivered using an HD-tES device (Soterix Medical, Inc, Woodbridge, NJ, USA), targeting the left dorsolateral prefrontal context (DLPFC). In the sham condition, the current was ramped up to 2mA within the first and last 30 seconds to mimic the sensation of stimulation, but then ramped down, with no current maintained at other times.

Assessments: Word-sentence association paradigm; Indicator of Negative interpretation bias.

Word sentence association paradigm (WSAP) was used to assess the interpretation bias in youth with social anxiety. Following the appearance of a fixed cross (500ms), an ambiguous sentence describing the social situation would be presented in the center of the screen (2000ms). Then, a word would appear at random for 500ms: the word might imply a threatening interpretation (e.g., "boring" or "rejection"), or imply a benign one (e.g., "captivating" or "accepted"). Participants were asked to decide whether the word was related to the ambiguous sentence ("F" for relevant and "J" for irrelevant).

This study recorded the response time and calculated the bias index (BI): threat BI = response times (reject threat - endorse threat) and benign BI = response times (endorse benign - reject benign).

Assessments: free recall task; Indicator of negative memory bias. This study used free recall and recognition task to assess memory bias by the number of positive/negative words remembered. First, in the "encoding phase", after a short time for a fixed cross, words were presented randomly in turn . Participants were asked to read these words aloud and answer coding questions.After this phase, there was a 3-minute distraction task, during which participants were guided through 40 calculation problems. They were then given 2 minutes to freely recall the words that had appeared before, regardless of the order of the words. Next, they completed a recognition task. Recall Accuracy was calculated as: (number of correctly recall negative words) ÷ (total number of negative words presented).

Assessments: the Liebowitz Social Anxiety Scale. The Liebowitz Social Anxiety Scale (LSAS) was used to identify participants with social anxiety and to assess changes in social anxiety symptoms before and after the tDCS intervention. The scale contains 24 items, each of which assesses two aspects of fear and avoidance, using a 4-point Likert scale ("0" means "never" and "3" means "severe"), for a total of 48 items. A higher score indicated more severe social anxiety. The score ranges from 0 to 144. The total score would be categorized into 5 severity levels: 0-54 (normal), 55-64 (moderate), 65-79 (marked), 80-94 (severe), and ≥95 (very severe). In this study, participants scoring between 55 and 95 were included. The LSAS has demonstrated good internal consistency across three assessments (Cronbach's alpha = 0.87-0.90), and acceptable test-retest consistency (r = 0.47-0.79, p < 0.001).

Assessments: the Spielberger Inventory - Trait (STAI-T). Trait anxiety was assessed using the Spielberger Trait Anxiety Inventory - Trait (STAI-T), which included 20 items rated on a 4-point scale, ranging from 20 to 80 (1 = "not at all" to 4 = "very much"). A higher score indicates severe trait anxiety symptoms. In this study, the STAI-T demonstrated good internal and test-retest reliability (Cronbach's alpha = 0.80-0.86; r = 0.69-0.79, p < 0.001).

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report attention bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the attention bias subscale includes 5 items (E.g., My attention is easily drawn to the tragic images on TV and is difficult to shift). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias. The score of attention bias subscale ranges from 5 to 25.

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report interpretation bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the interpretation bias subscale includes 5 items (E.g., If a new teacher is hard on me, I think it is because he sees me in a bad light and wants to get me in trouble). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias . The score of interpretation bias subscale ranges from 5 to 25.

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report memory bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the memory bias subscale includes 4 items (E.g., I can easily remember the negative comments people make about me). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias . The score of memory bias subscale ranges from 5 to 20.

Total of all reporting groups

Participants underwent 10 sessions of anodal (2mA, 20 minutes) HD-tDCS over 5 days (2 times each day) targeting the left dorsolateral prefrontal cortex (DLPFC).

Participants underwent 10 sessions of sham HD-tDCS over 5 days (2 times each day) targeting the left dorsolateral prefrontal cortex (DLPFC).

Assessment: Word-sentence association paradigm; Indicator of Negative interpretation bias.

Word sentence association paradigm (WSAP) was used to assess the interpretation bias in youth with social anxiety. Following the appearance of a fixed cross (500ms), an ambiguous sentence describing the social situation would be presented in the center of the screen (2000ms). Then, a word would appear at random for 500ms: the word might imply a threatening interpretation (e.g., "boring" or "rejection"), or imply a benign one (e.g., "captivating" or "accepted"). Participants were asked to decide whether the word was related to the ambiguous sentence ("F" for relevant and "J" for irrelevant).

This study calculated the proportion of threaten interpretations that participants endorsed as being related to the sentence （Endorsement rates of negative interpretations）, which was prescribed as the primary result of interpretation bias.

Assessments: Dot-probe task; Indicator of Attention Avoidance. Dot probe task is a common experimental paradigm for measuring attention bias. After appearance of a fixed cross in the center of the computer screen, a pair of faces appear side by side. Then, a probe appears on the location previously occupied by one of the two faces. Participants were asked to response the location of the probe as quickly as possible and press the appropriate keys to answer.

This study used the trail-level bias score (TL-BS) as a measure of the main outcomes of this task. TL-BS is an indicator of reaction time (RT), calculated by subtracting contiguous pairs of congruent trials (probe and emotional face appeared on the same location of the screen) from incongruent trials (probe and emotional face appeared on the opposite location of the screen).Negative scores represent avoidance of these stimuli. Peak TL-BSnegative indicates the minimum TL-BS value.

Assessments: Word-sentence association paradigm; Indicator of Negative interpretation bias.

Word sentence association paradigm (WSAP) was used to assess the interpretation bias in youth with social anxiety. Following the appearance of a fixed cross (500ms), an ambiguous sentence describing the social situation would be presented in the center of the screen (2000ms). Then, a word would appear at random for 500ms: the word might imply a threatening interpretation (e.g., "boring" or "rejection"), or imply a benign one (e.g., "captivating" or "accepted"). Participants were asked to decide whether the word was related to the ambiguous sentence ("F" for relevant and "J" for irrelevant).

This study recorded the response time and calculated the bias index (BI): threat BI = response times (reject threat - endorse threat) and benign BI = response times (endorse benign - reject benign).

Assessments: free recall task; Indicator of negative memory bias. This study used free recall and recognition task to assess memory bias by the number of positive/negative words remembered. First, in the "encoding phase", after a short time for a fixed cross, words were presented randomly in turn . Participants were asked to read these words aloud and answer coding questions.After this phase, there was a 3-minute distraction task, during which participants were guided through 40 calculation problems. They were then given 2 minutes to freely recall the words that had appeared before, regardless of the order of the words. Next, they completed a recognition task. Recall Accuracy was calculated as: (number of correctly recall negative words) ÷ (total number of negative words presented).

Assessments: the Liebowitz Social Anxiety Scale. The Liebowitz Social Anxiety Scale (LSAS) was used to identify participants with social anxiety and to assess changes in social anxiety symptoms before and after the tDCS intervention. The scale contains 24 items, each of which assesses two aspects of fear and avoidance, using a 4-point Likert scale ("0" means "never" and "3" means "severe"), for a total of 48 items. A higher score indicated more severe social anxiety. The score ranges from 0 to 144. The total score would be categorized into 5 severity levels: 0-54 (normal), 55-64 (moderate), 65-79 (marked), 80-94 (severe), and ≥95 (very severe). In this study, participants scoring between 55 and 95 were included. The LSAS has demonstrated good internal consistency across three assessments (Cronbach's alpha = 0.87-0.90), and acceptable test-retest consistency (r = 0.47-0.79, p < 0.001).

Assessments: the Spielberger Inventory - Trait (STAI-T). Trait anxiety was assessed using the Spielberger Trait Anxiety Inventory - Trait (STAI-T), which included 20 items rated on a 4-point scale, ranging from 20 to 80 (1 = "not at all" to 4 = "very much"). A higher score indicates severe trait anxiety symptoms. In this study, the STAI-T demonstrated good internal and test-retest reliability (Cronbach's alpha = 0.80-0.86; r = 0.69-0.79, p < 0.001).

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report attention bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the attention bias subscale includes 5 items (E.g., My attention is easily drawn to the tragic images on TV and is difficult to shift). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias. The score of attention bias subscale ranges from 5 to 25.

Assessments: recognition task; Indicator of Negative Memory bias. This study used free recall and recognition task to assess memory bias by the number of positive/negative words remembered. First, in the "encoding phase", after a short time for a fixed cross, words were presented randomly in turn . Participants were asked to read these words aloud and answer coding questions.After this phase, there was a 3-minute distraction task, during which participants were guided through 40 calculation problems. They were then given 2 minutes to freely recall the words that had appeared before, regardless of the order of the words. Next, they completed a recognition task. Recognition Accuracy was calculated as: (number of correctly recognized negative words) ÷ (total number of negative words presented).

Please note that a total of 47 participants completed the memory bias tasks (25 in the tDCS group, and 22 in the sham group).

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report interpretation bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the interpretation bias subscale includes 5 items (E.g., If a new teacher is hard on me, I think it is because he sees me in a bad light and wants to get me in trouble). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias . The score of interpretation bias subscale ranges from 5 to 25.

Assessments: the Negative cognitive processing bias questionnaire; Indicator of self-report memory bias.

Participants completed the Negative cognitive processing bias questionnaire (NCPBQ). The NCPBQ included 17 items divided into three subscales, along with three lie-detection items. Among it, the memory bias subscale includes 4 items (E.g., I can easily remember the negative comments people make about me). Items were rated on a 5-point Likert scale from 1 ("disagree completely") to 5 ("agree completely"), with higher score indicating more severe cognitive bias . The score of memory bias subscale ranges from 5 to 20.

Participants underwent 10 sessions of sham HD-tDCS over 5 days (2 times each day) targeting the left dorsolateral prefrontal cortex (DLPFC).