Table 3.

Computational considerations of mental health CAs^a for youth (N=39).

Dimensions and codes				Value, n (%)		Key trends
System characteristics						Most of the mental health CAs were prototypes built upon existing mobile chat apps
	Maturity of device
		Prototype	26 (67)
		Fully functioning system	7 (18)
		Existing system	8 (20)
		Concept	1 (3)
	Delivery channel
		Mobile	16 (41)
		Web	11 (28)
		Desktop	6 (15)
Communication mode						All CAs supported text-based input (free text along with quick options) with a few that supported voice input; all supported textual output, with many of them supporting visualized output such as image and video
	Input mode
		Text	39 (100)
		Text+speech	2 (5)
		Free text+options	19 (49)
		Free text	13 (33)
		Quick options	5 (13)
	Output mode
		Text	39 (100)
		Image	21 (54)
		Video	12 (31)
		Audio	7 (18)
		Game	6 (15)
	AIb technique					Free-textual inputs were processed via NLPc, while prewritten content was delivered via rule-based programming
		NLP	19 (49)
		Rule based	15 (38)
	Content delivery					Most mental health content was delivered in flexible ways; in some cases, prewritten content was delivered in a structured manner
		Flexible	25 (64)
		Structured	8 (20)
		Semistructured	3 (8)
	Knowledge base					Mental health content was built upon evidence-based expert knowledge of cognitive and behavioral therapy and positive psychology
		CBTd	19 (49)
		Positive psychology	7 (18)
		Other therapeutic content	7 (18)
		Clinical expert knowledge	6 (15)

^aCA: conversational agent.

^bAI: artificial intelligence.

^cNLP: natural language processing.

^dCBT: cognitive behavioral therapy.