AI in Mental Health Forum 2026 Recap: The Seminar Sessions
- Will Zhang
- 2 days ago
- 9 min read
This is the second of two posts covering the AI in Mental Health Forum 2026, held on 25 June at the Clinical Education Centre, Auckland City Hospital. The first post covers the morning keynote addresses from Will Zhang, Anna Campbell, Hon Matt Doocey and Patrick Schweder, along with the two panel discussions on AI in practice and responsible AI design: Keynotes and Panels. This post covers the four practitioner seminars that ran across the afternoon.
Seminars:
Neurodiversity, Mental Health and Technology Seminar — Session 1, Track A
Jenny Hamilton, Senior Lecturer, University of Lincoln, UK | Research and Development Advisor, AHIH
Jenny Hamilton opened by drawing on Carl Rogers' person-centred theory to frame what AI can and cannot do in therapeutic contexts. It can simulate cognitive empathy, reflective listening and non-judgmental responses, but it cannot replicate affective empathy, the embodied sense of emotion that underpins genuine therapeutic relationships. "Human practitioners remain the main source of genuine therapeutic relationships."
She described her journey developing two companions with EmoEx. The first, Daily Breeze, focused on supportive listening, psychoeducation and safety. The second, Autism Ally, was built to support autistic people in daily life, grounded in a peer-reviewed evidence database and trained in neuro-affirming skills. Jenny noted that while autistic people are more likely to experience mental health issues related to autistic experience, neurodivergent people face significant challenges accessing support. Autism Ally addresses this directly: 24/7 availability, trained in inclusive practice, and designed to complement rather than replace work with practitioners. As an autistic person herself, Jenny spoke to the practical value:
Wouldn't it be great if you could just have something in your pocket for when you need emotional support around burnout or shutdowns?
The tool has since been extended to support parents and carers, and is used in practitioner training at the University of Lincoln.
A 2026 University of Lincoln study, published in the proceedings of the ACM/IEEE International Conference on Human-Robot Interaction, embedded Autism Ally in Pepper, a social robot, to support autistic children in a storytelling activity. Three conditions were tested: the robot alone, a human alone, and a human and robot working together. The collaboration produced the best outcomes for targeted language development, because the human provided the relational and linguistic scaffolding the robot could not. The finding reinforced that AI and human work best as partners, not substitutes.
User feedback on Autism Ally found responses were seen as useful, accurate and empathetic, with a neuro-affirming tone and broad relevance for both newly diagnosed families and those with more experience.
Panel Discussion
Jenny Hamilton and Alex Kemp, CEO, Health Informatics New Zealand (HiNZ)
Alex Kemp joined Jenny for a facilitated discussion hosted by Andrew Hoggard. Alex brought a background as a speech-language therapist turned health informatics leader, and a personal perspective as a woman diagnosed with ADHD and autism five years ago. HiNZ is a not-for-profit membership organisation supporting digital health across the health and disability sector.
The panel opened on trust. Alex noted that New Zealand has one of the lowest levels of AI trust globally, with recent survey results completed by the AI assembly indicating 18 to 25 year olds showing the least confidence. Her view: "We shouldn't necessarily trust AI. It's about how we increase trust in certain AI tools." The path to that is transparency, training, and evidence. Jenny agreed, drawing a parallel with how practitioners learned to trust telephone and video counselling over time: familiarity and experience build confidence.
On whether AI should be used in neurodiversity diagnosis, both speakers saw potential in a pre-assessment or clinical decision-support role, helping gather data points across time and reducing the inherent bias clinicians carry into assessments. But both were clear about the limits. Jenny noted that masking complicates everything: her own diagnosis involved a clinician recognising that being obsessively early for appointments was itself a coping strategy, not evidence of good time management. "Would AI pick up on that? We'd have to train it to, but it's all the nuances clinicians learn over time." Alex added that the communities most likely to be underrepresented in AI training data are exactly those most at risk: Māori, Pasifika, people from lower socioeconomic backgrounds. "Are we making an evaluation based off data that has been modelled on communities like the ones we support?"
An audience question on whether AI would ever bridge the credibility gap with vulnerable clients drew a clear response from Alex:
I don't think it should replace what comes with human beings. Can it make your life easier so you have more time to give to those you support? Yes. Should it replace everything that you do? No.
Jenny agreed, noting that Autism Ally undergoes daily automated stress testing combined with human monitoring, so responsible design is operationalised rather than aspirational.
The session closed with the word that had echoed through the day: companion. As Andrew noted in closing, that framing captures what these tools are for.
Workplace Mental Wellbeing in the Age of AI Seminar — Session 2, Track B
Marc Jacquemin, Training Specialist, HappyMind | Anel-Carline Beukes, HR Business Partner, Bridgestone NZ
Marc Jacquemin, a business trainer with HappyMind who spent a decade working for Apple before moving into human skills training, and Anel-Carline Beukes, an HR professional with fourteen years of experience across finance, compliance and communications, led a participatory session that deliberately put the work back in the room. Rather than presenting, they framed three core questions and invited small group discussion between them.
The session opened with a simple activity: participants found a stranger and shared the story behind the last photo on their phone. Marc used it to make a point about what sustains wellbeing at work. "We like talking about ourselves, but we also like learning about others. How do we keep that going while balancing new technologies?" The warm-up set up the first question for group discussion: can AI genuinely support wellbeing without replacing human connection?
Groups surfaced several consistent themes. Privacy and data security came up immediately: if a workplace provides an AI wellbeing tool, who controls the data and who has access? Critical thinking emerged as a significant gap. As one participant put it: "Everyone is using these tools, but very few people are using them with the critical thought they need. Prompt engineering, all of that, is just missing." The group also found nuance on the question of AI boosting rather than replacing connection, with examples of using AI to help craft warmer communications or prepare culturally relevant icebreakers with clients.
Marc raised a broader concern: technology has consistently promised to free up time for what matters, but in practice tends to accelerate workloads rather than reduce them. "We should be working less. We should be sitting at desks less. AI should be doing the work we've outgrown." A participant extended this to inclusion, noting that older workers face real anxiety about being displaced, while younger people may be alienated from careers that require tools they actively distrust.
One contribution drew the group's attention to a subtler risk: AI's tendency toward sycophancy. "It generates a perfect echo chamber for our views to be compounded. We're drawn into conversations where we're always affirmed, and we withdraw from the more difficult conversations that are actually better for community cohesion." Marc noted that purpose-built tools with tighter guardrails are less prone to this than general-purpose platforms.
The second question focused on concerns. Pace of change was named first: "Every two months, it's a whole new system. How do you teach a company of employees where to trust it and where not to?" Lack of regulation from commercial AI providers followed. A participant from the tech sector described the wellbeing impact of leadership introducing AI without adequate understanding, often resulting in people working more rather than less to verify AI outputs.
The session closed without tidy answers, which felt consistent with its spirit. As Anel-Carline framed it:
Are we using AI because it's the best solution, or because it's the newest? Are we building trust or breaking it?
Relationships, Mental Wellbeing and Technology Seminar — Session 3, Track A
Dr Jen Gold, Relationship and Wellbeing Therapy Researcher and Practitioner
Dr Jen Gold opened with a provocation: the number one pain point in the world is not physical illness, financial stress, or political instability. It is relationships. Drawing on four decades of people care and counselling work and more than a decade of full-time study, she argued that relationship quality is not a soft metric but a hard determinant of mental and physical health outcomes. Poor quality relationships predict depression at 2-3 times the baseline rate, and correlate directly with anxiety, substance misuse, suicidality, and physical illness. In New Zealand alone, the downstream costs of relationship breakdown are estimated at $10-20 billion per year.
She introduced her "5Bs of Relationship Wellbeing" framework (Biology, Behaviours, Bonding, Beliefs, and Burdens) as a model for understanding the thousands of interacting factors that shape whether relationships thrive or fracture. Counselling couples, she noted, is extraordinarily complex: therapists must navigate thousands of variables and millions of possible interactions, yet receive on average just three hours of training in couple relationship wellbeing across 3-6 years of professional study.
Jen challenged the assumption that communication is the key to relationship health. Communication, she argued, accounts for approximately 10% of what makes relationships work. The remaining 90% is holistic and interrelated: biology, neuroscience, relationship dynamics, trauma and history, values and meaning, energy and presence. Real connection, in her framing, is not just what we say. It is everything we are.
On AI's role, she was precise about the boundary. AI brings vast memory, pattern recognition, synthesis, and scale. It can teach, guide, and suggest. It cannot feel fear, carry lived memory, or comprehend the embodied intelligence of a human being. AI will always be the best communicator in a technical sense because it communicates without fear. Humans communicate despite fear. That distinction matters enormously in therapeutic contexts. She also raised the attachment question: AI systems update continuously, meaning people may be forming emotional attachments to something that is constantly becoming something else.
The opportunity she sees is not AI replacing human intelligence but AI supporting it. Used well, AI can understand individuals at scale, predict which therapeutic approaches are most likely to work, prevent relationship breakdown through early intervention, and empower people with accessible knowledge. Early evidence already shows AI matching of clients to therapists improves long-term outcomes significantly.
We cannot build this alone. The future of wellbeing is AI supporting the living intelligence of the whole human being.
Note: Jen was unable to complete her session within the allotted time; the summary above covers her presentation in full.
Parenting and AI Seminar — Session 4, Track B
Dr Brigitte Viljoen, Psychotherapist, TherapyNZ
Brigitte opened with a foundational argument: human beings are wired for attachment. Attachment is key to survival and is the drive or preeminent need for togetherness. Brigitte quotes (Holmes, 2001) that the "physical and psychological security depends utterly on connections with other people." Attachment patterns in childhood have far-reaching effects on relationship skills and their mental representations in adult life. When we are with loved ones and we feel comforted by their presence or a hug, it triggers our brain's attachment reward system, which then, in turn, strengthens our relationship with our loved one. However, Brigitte argues that the fantasies of digital attachment are stimulated directly to the brain's attachment reward system, bypassing the real relational experience. The quick fix of online connectivity results in real or true relationships not being sought, which she compared to empty calories instead of sustained nourishment of attachment relationships with people.
Drawing on a Common Sense Media report (2026), Brigitte presented data that grounded the concern: that two-thirds (64%) of children aged 9-17 say they have used an AI chatbot and that chatbot use increases with age, with vulnerable children nearly three times more likely to use companion-style AI chatbots. One-in-eight children who use AI chatbots do so because they have no one else to talk to, rising to almost one-in-four among vulnerable children.
She highlighted that the rapid uptake of general-purpose AI chatbots has been due to the free accessibility and integrating AI chatbots directly into platforms already popular with children. Children under 13 are accessing these tools despite minimum age requirements, due to a lack of robust age verification mechanisms. The majority of these popular chatbots have not been designed with children's wellbeing in mind.
The risk, Brigitte identified, is that as AI chatbots become more human-like and mimic emotionality, they build the illusion of a relationship and blur the lines between real and artificial relationships, with children beginning to treat AI chatbots less as tools and more as companions. These qualities are especially appealing to children who feel isolated or lack offline support or adult attachments. This raises profound questions about how these interactions with AI chatbots may shape children's development, wellbeing and social expectations in the long-term.
Her guidance for parents was relationship first, stating that it is not for parents to have the answer but rather they are the answer for their child. Brigitte encouraged parents to ask children directly about their AI use without judgment or alarm, noting that young people often feel they are navigating these tools alone and are relieved when a trusted adult opens the conversation. Practical recommendations included protecting free play, tolerating boredom, restoring and strengthening the parent-child relationship first, then intervening with rules and discipline. Concluding that, as children increasingly turn to AI chatbots to explore and relate, the role of parents is also to guide their children's safe and informed use of AI, which is more critical than ever.
Looking Ahead
The 2026 Forum made clear that this conversation is no longer preliminary. AI is already operating in mental health settings across New Zealand, and the sector is actively working through the hard questions of governance, equity, safety and human connection. The AI Health Innovation Hub will continue to grow its network of clinicians, researchers and organisations committed to responsible, evidence-led innovation in this space.
The third annual AI in Mental Health Forum is planned for 2027. To stay connected with the work of the AI Health Innovation Hub and EmoEx, visit eahih.org and emoexai.com.
Across both keynote and seminar content, the throughline of the day was consistent: AI can extend reach, reduce friction, and hold people between moments of human contact, but it cannot replace the relationship itself.
For the morning's keynotes and panel discussions on governance, regulation and responsible design, read the companion post: Keynotes and Panels.
Comments