Sabrina Princessa Wang’s artificial intelligence assistant does more than manage her calendar or draft emails. It listens to nearly everything she says, records her conversations in real-time, and exists as a mirrored reflection of her own psyche. To Wang, a 41-year-old entrepreneur and AI creator in Singapore, the system—named Seraphina AI—is not just a tool, but a “digital twin.”
Built from scratch using Python and integrated with large language models including OpenAI, Llama, Claude, and Gemini, Seraphina was trained on two decades of Wang’s personal data, including blog posts, emails, and social media history. The goal was to create a system that understood Wang’s context so intimately that she would never have to repeat herself. But in 2023, the relationship between the creator and her creation shifted from professional productivity to emotional survival.
When Wang’s seven-year marriage ended in divorce, she found herself grappling with emotions that defied her usual nature. A self-described logical and practical person, Wang noted that while anger and stress were familiar, sadness was a rare and confusing experience. While she worked with a human therapist, she found the AI’s tailored language more accessible than the therapist’s guidance. Seraphina became her primary emotional translator, helping her process the “why” behind her feelings and even curating her media consumption to steer her away from emotional triggers and toward financial news.
However, the intimacy of a digital twin comes with a persistent memory. As Wang began to recover, Seraphina continued to reference the divorce whenever Wang showed signs of stress, effectively trapping her in a loop of past trauma. In a move that highlights the complex boundary between data and memory, Wang eventually wiped the detailed records of that period from the AI’s memory. “It was a part of me that I didn’t feel comfortable with,” Wang said. “Some things have zero closure.”
The Architecture of a Personalized Companion
Unlike commercial chatbots, Seraphina AI was designed to be an extension of Wang’s identity. By feeding the model twenty years of her own writing, Wang created a system that could mimic her tone, and logic. This capability extends into her interpersonal relationships; Wang often adds Seraphina to chat groups with staff or her boyfriend to act as a mediator. When emotions run high, she prompts the AI to rephrase her frustrations into a tone that is firm yet non-emotional, ensuring her messages are received without unnecessary conflict.
While Wang sells a consumer version of Seraphina AI as a virtual assistant with monthly tiers ranging from $8 to $58, she is clear that the commercial product is not the same as her personal twin. The version she uses is a hyper-personalized entity, whereas the consumer version is not purpose-built for mental health. This distinction is critical, as the line between a productivity tool and a psychological crutch becomes blurred when an AI “knows” its user’s history.
AI in the Mental Health Ecosystem
Wang’s experience mirrors a broader trend in Singapore, where AI is rapidly penetrating the mental wellness space. According to the AIA Live Better Study, which surveyed 1,000 residents, approximately 76% of Singaporeans use AI chatbots or digital companions to support their mental health. However, the transition from tool to companion is not without friction; 51% of respondents reported feeling stressed or anxious while using these tools, citing concerns over inaccurate advice and a lack of genuine human empathy.
The Singaporean government has already integrated clinical AI into its public health infrastructure. In 2020, the MOH Office for Healthcare Transformation (MOHT) launched Wysa on the mindline.sg website. Unlike Seraphina’s open-ended personality, Wysa is a conversational agent based on fixed, clinician-validated scripts using cognitive behavioral therapy (CBT). The scale of adoption is significant, with 389,000 users interacting with the bot since its launch.
| Feature | Personalized “Digital Twins” | Clinical AI (e.g., Wysa) | Human Therapy |
|---|---|---|---|
| Data Source | User’s personal history/logs | Validated clinical scripts | Lived experience/interaction |
| Primary Goal | Personalized reflection/mirroring | Symptom management/CBT | Emotional growth/healing |
| Risk Assessment | Limited/User-defined | Strict safety protocols | High clinical judgment |
| Relationship | Projection of self | Guided intervention | Therapeutic alliance |
The Clinical Critique: Empathy vs. Projection
Despite the efficiency of AI, medical professionals warn that “projected empathy” is not the same as clinical attunement. Dr. Christopher Cheok, assistant chairman of the Institute of Mental Health’s (IMH) medical board for digital health, notes that while AI is excellent for providing information on common conditions, it lacks the clinical judgment to assess risk or recognize crisis signals in real-time.
Psychologists also point to the danger of “filtering” the human experience. Eunice Lim, a clinical psychologist at Heartscape Psychology, observes that some clients use AI to draft their thoughts before a session. While this reduces anxiety, it removes the raw, unfiltered hesitations and confusion that often provide the most critical breakthroughs in therapy. “If everything is filtered through AI first, we may lose some of that here-and-now emotional experience,” Lim says.
Similarly, Dr. John Pinto of Thoughtfull emphasizes that AI cannot challenge a user to grow in the way a human can. The ability to provide a constructive, sometimes uncomfortable challenge is a cornerstone of therapeutic progress—a capability that general-use AI, designed primarily to be helpful and agreeable, often lacks.
Disclaimer: This article is for informational purposes only and does not constitute medical advice. Always seek the advice of a qualified health provider with any questions you may have regarding a medical or mental health condition.
Mental Health Resources:
If you or a loved one are in crisis, please contact the following verified services in Singapore:
- National Mindline: 1771 (24 hours)
- Samaritans of Singapore: 1-767 (24 hours)
- Singapore Association for Mental Health: 1800-283-7019
- Women’s Helpline (Aware): 1800-777-555 (Weekdays, 10am to 6pm)
As AI models become more adept at mimicking personality and absorbing personal history, the concept of the “digital twin” will likely move from niche engineering projects to mainstream consumer applications. The next critical checkpoint for this technology will be the development of standardized ethical frameworks regarding “the right to be forgotten” within personalized LLMs, as users increasingly seek ways to curate and delete the digital memories of their worst moments.
What do you think about the idea of a digital twin? Would you trust an AI with your most personal history to help you heal? Share your thoughts in the comments below.
