Scientific validation of conversational AI systems for emotional wellbeing: results of the Blind Echo Experience program

Executive summary

Objective: To assess the effectiveness, acceptability and commercial viability of Somia™ (conversational AI system) and VVAVVE™ (wellness mask) in emotional regulation, using the Blind Echo Experience protocol.

Methods: A six-month pilot study at The Terminal Hub (n=89 participants, 119 sessions). A structured six-phase protocol including an immersive experience, conversational interaction and multi-metric assessment. The preparation of the experimental setting and the methodological design were supervised by the Chief Science Officer, Dr David Martínez Rubio. The initial evaluation of results involved public-private research groups.

Results: 83.1% reported a significant improvement in perceived stress after the intervention. 76.4% reported an increase in mental clarity. 97.8% rated the experience as satisfactory. 58.7% of decision-makers expressed an intention to purchase. 23 users completed recurring sessions with no evidence of habituation.

Conclusions: The Somia™-VVAVVE™ combination demonstrates greater effectiveness than commercial mindfulness applications, with high acceptability. The results support progression to ERL-3 for commercial transfer under frameworks of ethical responsibility.

1. Introduction and context

Artificial intelligence technologies applied to emotional wellbeing have grown exponentially, reaching market valuations above USD 5.6 billion. Most, however, lack rigorous scientific validation, with independent studies reporting effectiveness rates varying between 38% and 64% for commercial mindfulness applications.

This gap between technological innovation and empirical evidence raises fundamental ethical challenges. Deploying conversational AI systems in emotional wellbeing contexts without adequate validation can create false expectations or unanticipated adverse effects.

yeshcube has developed an integrated ecosystem combining Somia™ (conversational AI specializing in emotional support) with VVAVVE™ (a wellness mask with sensory isolation and 3D spatial audio). These systems were designed under principles of algorithmic empathy, privacy by design and progressive scientific validation according to the Evidence Readiness Level (ERL) methodology.

Research objectives

Primary objective: To assess the effectiveness of the Somia™-VVAVVE™ integration in reducing perceived stress and improving mental clarity.

Secondary objectives: To determine technological acceptability, commercial viability and the safety profile in a diverse professional population.

2. Methodology

2.1 Experimental design

A prospective pre-post pilot study without a control group, run over six months (January–June 2025) at specialist facilities in The Terminal Hub, Valencia. The Blind Echo Experience protocol was designed to evaluate emotional regulation systems under controlled conditions while retaining ecological relevance.

2.2 Participants and sampling

Inclusion criteria: Adults aged 18 or over, able to give informed consent, available for a full session (45–60 minutes).

Exclusion criteria: History of severe claustrophobia, active psychotic episodes in the previous 12 months, use of hearing devices incompatible with the wellness mask.

Sample characteristics:

Variable	n	%
Total participants	89	100.0
Sessions completed	119	—
Gender distribution
Women	58	65.2
Men	31	34.8
Age distribution
18–25 years	9	10.1
26–50 years	62	69.7
>50 years	18	20.2
Professional profile
Executives/decision-makers	49	55.1
Operational professionals	18	20.2
Middle management	13	14.6
Other profiles	9	10.1

2.3 Technological intervention

VVAVVE™: A wellness mask with total light isolation, a 3D spatial audio system and wireless connectivity. Pre-recorded experiences of variable duration (Care Program for Work/Life).

Somia™: A conversational AI system with a multi-layer prompting architecture for emotional support contexts. Algorithmic transparency and safe redirection protocols built in.

2.4 Experimental protocol

Phase	Duration	Activity
1	1 min	Instructions and consent
2	7 min	VVAVVE™ immersive experience
3	5–10 min	Somia™ conversational interaction
4	1 min	Usability assessment
5	1 min	Emotional impact measurement
6	1 min	Commercial validation

Average total duration: 18.3 minutes (range: 16–22 min)

2.5 Variables and measurements

Primary variables:

Perceived stress: 1–10 Likert scale (pre- and post-intervention)
Mental clarity/focus: 1–10 Likert scale (post-intervention)
Overall satisfaction: 1–5 Likert scale

Secondary variables:

Qualitative descriptors of the experience
Frequency of recurring use
Willingness to recommend institutionally
Intention to adopt commercially

2.6 Statistical analysis

Descriptive statistics for categorical and continuous variables. Calculation of 95% confidence intervals. Effect size analysis (Cohen’s d) for pre-post changes. Thematic content analysis for qualitative data.

2.7 Ethical considerations

Approval by yeshcube’s Ethics Committee. Verbal and written informed consent. Interruption protocols available. Full anonymisation of personal data.

3. Results

3.1 Baseline characteristics and completion

89 participants completed the full experimental protocol (completion rate: 100%). 23 participants (25.8%) took part in multiple sessions during the study period, for a total of 119 experimental sessions.

Distribution of experience modes:

Care Program for Work: 54.6% (n=65)
Care Program for Life: 45.4% (n=54)

Mean duration of the immersive experience: 6.8 minutes (SD=1.2; range: 5.1–8.9 min). Zero drop-outs during active sessions.

3.2 Effectiveness in emotional regulation

Reduction in perceived stress:

Metric	Value	95% CI
Participants with significant improvement	74/89 (83.1%)	75.2–89.1%
Mean change on Likert scale	-3.4 points	-3.8 to -3.0
Effect size (Cohen’s d)	1.52	1.23–1.81

Improvement in mental clarity: 68 participants (76.4%) reported an increase in mental clarity, calm or capacity to focus. Mean post-intervention score: 7.9/10 (SD=1.4).

3.3 Acceptability and user experience

Overall satisfaction: 87 participants (97.8%) rated the experience as satisfactory (≥4/5 on the Likert scale). Mean score: 4.6/5 (SD=0.6).

Distribution of qualitative descriptors:

Descriptor	n	%
“Surprising”	79	88.8
”Striking”	43	48.3
”Intense”	11	12.4
”Slightly disappointing”	6	6.7

Indicators of future adoption:

Would recommend workplace implementation: 96.6% (86/89)
Would use AI for emotional support: 94.4% (84/89)
Willing to repeat the experience: 78.7% (70/89)
Interested in follow-up through an app: 61.8% (55/89)

3.4 Commercial validation

Willingness to adopt at organizational level: Among the 49 participants with organizational decision-making capacity, 28 (57.1%) expressed a willingness to acquire similar systems.

Behavioral evidence of commercial interest:

Letters of institutional support: 23
Letters of specific purchase intent: 9
Requests for detailed technical information: 31

3.5 Analysis of recurrence and sustainability

23 participants completed multiple sessions (mean: 2.7 sessions per recurring user).

Effectiveness in recurring users:

Benefits maintained: 91.3% (21/23)
No habituation: sessions-effectiveness correlation r=-0.12 (p=0.58)
Increase in satisfaction with repeated use: 17.4% (4/23)

3.6 Safety profile

Safety-related events: Zero adverse events reported. Zero episodes of claustrophobia or increased anxiety during sessions.

Interruptions: 1 participant (1.1%) requested an early finish out of personal preference, unrelated to safety.

4. Discussion

4.1 Comparative effectiveness

The results obtained (83.1% improvement in perceived stress) significantly exceed the ranges documented for commercial mindfulness applications (38–64%) and approach the effectiveness reported for structured in-person interventions (78–89%).

The calculated effect size (d=1.52) indicates robust clinical significance, suggesting a potentially measurable impact on organizational indicators of occupational wellbeing.

4.2 Contributing factors

Multimodal integration: Combining sensory isolation (VVAVVE™) with adaptive conversational guidance (Somia™) may produce synergistic effects greater than either modality alone.

Contextual personalization: Dynamic adaptation algorithms allow relevance to be optimized according to the specific context and the user’s responses.

Design principles: Implementing the Audio-first and Zero-screens paradigms lowers adoption barriers compared with more complex interfaces.

4.3 Commercial viability

The willingness to adopt (57.1% among decision-makers) exceeds typical early adoption rates for emerging organizational technologies (18–28%). The behavioral evidence (letters of support and purchase intent) reinforces the commercial viability documented quantitatively.

4.4 Methodological limitations

Experimental design: The absence of a control group limits definitive causal attribution of the effects observed. Self-report as the primary measure introduces potential social desirability bias.

Sample characteristics: The concentration of urban professional profiles limits generalization to populations with different demographic characteristics.

Time frame: An assessment limited to immediate effects provides no information about how durable the benefits are.

4.5 Directions for future research

Controlled studies: Implementing designs with random assignment to experimental versus active control conditions.

Objective measures: Incorporating physiological biomarkers (heart rate variability, cortisol) to validate subjectively reported changes.

Longitudinal follow-up: Assessment at 3–6 months to determine whether the benefits are sustained.

Population diversification: Extension to clinical contexts and different demographic groups.

5. Conclusions and implications

5.1 Validation of effectiveness

This study provides empirical evidence about the effectiveness of integrated conversational AI systems and wellness masks for emotional regulation in a professional population. The results exceed documented standards for similar technologies in effectiveness, acceptability and commercial viability.

5.2 Responsible transfer

The findings support progression toward ERL-3 under yeshcube’s methodology: evidence of effectiveness in a real context, verified acceptability, a favorable safety profile and documented commercial viability.

5.3 Contributions to knowledge

The research sets methodological precedents for validating emotional wellbeing technologies and provides evidence about the potential to democratise access to scientifically validated interventions.

5.4 Practical applicability

The results suggest applicability in workplace, educational and community contexts, with potential for integration into organizational health and wellbeing policies.

References

[1] Grand View Research. (2024). Mental Health Apps Market Size & Growth Report.
[2] Flett, J.A.M., et al. (2019). Mobile mindfulness meditation: a randomized controlled trial. Journal of Medical Internet Research, 21(3), e12841.
[3] Bennike, I.H., et al. (2017). Internet-based mindfulness training increases well-being and decreases stress. Journal of Health Psychology, 22(8), 1042-1052.
[4] Baumel, A., et al. (2017). Digital mental health interventions: A systematic review. Internet Interventions, 9, 1-13.
[5] Goyal, M., et al. (2014). Meditation programs for psychological stress and well-being. JAMA Internal Medicine, 174(3), 357-368.

Correspondence: Carolina Cachero, carol.cachero@yeshcube.com

Funding: This research was carried out with yeshcube’s own resources in collaboration with The Terminal Hub.

Data availability: The data from this research is available on request for academic research purposes. Market validation data is available on specific request, subject to data use and privacy protection agreements.

Would you like to collaborate, co-research or explore our lines of work?

Discover our Allies program →

Visit the Somia website →

Visit the VVAVVE website →

Fill in our contact form →