EXECUTIVE SUMMARY
This is an AI
assisted brief for software developers. It asks for proposals to create a 'clinical
decision-support and self-management tool' that can meet the growing
demands for complex, high-risk case management, being placed on primary care mental health services. This is particularly acute where NHS functions overlap those of the Home Office (e.g. controlled drugs, domestic violence, asylum seekers, fixated threat, and sex offenders).
The proliferation of smart technologies is shifting the risk landscape from public meeting places to domestic addresses. This has become colloquially known as the 'attention economy'. A digital
therapeutic platform is required to help people regain voluntary control over
their attentional resources in response to psychologically salient cues
associated with addiction, with or without any comorbid anxiety, depression or other psychiatric or neuroatypical conditions.
The software should operationalise market-leading evidence-based psychological principles (CBT, motivational interviewing, relapse prevention, attentional training, and mindfulness-based strategies) and empower service-users' recovery from addictions. Rather than treating symptoms after they develop, the platform should aim to intervene in earlier and earlier phases of the cognitive processes associated with an addiction, up to the point where emotionally significant cues capture attention.
Thus, in theory, the software should assist users in recognising attentional capture, understand its relationship to their emotional states (including trauma and attachment histories), identify and tolerate unmet needs and cravings, and assist in the practising of alternative attentional and behavioural responses until these have become habitual. The platform is intended to complement existing NHS Talking Therapies, Primary Care Mental Health Teams and addiction services rather than replace them.
CLINICAL RATIONALE
The therapeutic target is not necessarily the low mood or even the
addictive behaviour itself, but the person's ability to tolerate cues and cravings
without acting on them. This may have previously been referred to as inhibitory control or response
inhibition in the literature.
The basic ideas
involved have strong support from several therapeutic traditions:
·
CBT
teaches people to identify triggers and test alternative responses.
·
Motivational
Interviewing strengthens commitment to change.
·
Relapse
Prevention focuses on anticipating high-risk situations.
·
Acceptance
and Commitment Therapy (ACT) teaches willingness to experience cravings without
acting on them.
·
Mindfulness-Based
Relapse Prevention explicitly trains people to "surf" urges until
they pass.
·
Marlatt’s
"urge surfing" refers to learning that cravings are temporary
internal events that do not require action.
Although these therapeutic principles are well established, meta-analyses generally report only small-to-moderate average treatment effects, with considerable heterogeneity between studies and substantial relapse rates. Several authors have argued that this variability may reflect inadequate characterisation of individual differences, including impulsivity, executive control, neurodevelopmental traits and other baseline cognitive characteristics (re: Moeller et al 2015;
Burton
et al. 2021; Mehta et al
2021; Fascher
et al 2023; Thai
et al 2024; Di
Rosa et al 2026).
Thus, this call
is for software to be developed that is based on the proposition that:
1)
Many
common mental health difficulties involve repeated reinforcement of attentional
biases towards emotionally significant cues.
2)
Attentional biases appear to be modifiable independently of underlying attachment histories, trauma exposure and neurodevelopmental traits, although the extent to which such changes translate into durable clinical improvement remains an active area of research.
3) Repeated reinforcement strengthens associative links between emotionally salient cues, expectations and behavioural responses. Through spreading activation within associative memory networks, these strengthened associations increase attentional capture, craving, rumination and expectations of future reward.
4)
The
therapeutic objective is therefore to strengthen voluntary attentional control
before behaviour is initiated.
5)
Repeated successful disengagement from maladaptive cues is hypothesised to strengthen inhibitory learning, reduce the functional salience of those cues and strengthen adaptive behavioural responses.
Primary
Clinical Objectives
The software should enable users to: recognise emotional vulnerability; identify high-risk internal and external cues; detect early attentional capture; interrupt escalating craving or rumination; practise alternative attentional responses; reinforce successful self-regulation; develop long-term resilience.
THE PROPOSED USER JOURNEY
Phase 1 –
Personal Learning
The
objective is to build an individual attentional profile. The software learns:
·
situations
associated with distress
·
times
of day
·
locations
·
emotional
states
·
people
·
activities
·
digital
behaviours
·
language
associated with cravings.
Phase 2 –
Cue Recognition
The
software teaches users to notice: "I've become drawn towards
something" before "I need this." Exercises focus on recognising:
·
bodily
sensations
·
thoughts
·
emotional
shifts
·
attentional
narrowing.
Phase 3 –
Attentional Reallocation
Once
cues are recognised, the software guides users through evidence-based exercises
such as:
·
attentional
shifting
·
urge
surfing
·
cognitive
defusion
·
motivational
interviewing prompts
·
breathing
exercises
·
behavioural
substitution
·
value-based
decision making.
Phase 4 –
Reinforcement
Every
successful interruption is positively reinforced. The emphasis is on mastery
rather than symptom scores. The software visualises:
·
cravings
successfully interrupted
·
time
spent disengaging
·
confidence
·
resilience
growth
·
reduction
in automatic responding.
ARTIFICIAL INTELLIGENCE COMPONENT
The
AI should personalise rather than diagnose and:
Detect
recurring patterns in:
· emotional states
· environmental contexts
· cue exposure
· behavioural sequences.
Personalise
interventions according to:
· previous success
· time of day
· motivation
· current emotional state
· historical relapse patterns
· periods of elevated vulnerability
· preferred pathways of support
Provide
clinicians with a clinical dashboard of:
· personalised cue maps
· relapse trajectories
· resilience measures
· engagement statistics
· intervention effectiveness
· patient-defined goals.
Be safe and secure enough to:
· comply with NHS data governance requirements
· explain every AI recommendation
· allow patients full control over data sharing
· store the minimum necessary personal information
· avoid commercial advertising or attention-maximising design.
DESIGN PHILOSOPHY
Unlike
commercial digital platforms that compete for user attention, this platform
should be explicitly designed to strengthen users' capacity to regulate,
control, and redirect attention voluntarily.
Success is
measured not by time spent using the application but by increasing independence
from it.
Most digital
health apps ask:
"How are you feeling today?"
This theory
suggests a different primary question:
"What has captured your attention today?"
That is a
subtle but profound shift.
Mood becomes a
consequence of attentional allocation rather than the only thing being
measured.
The app could
build an attentional diary, not just a mood diary.
Imagine a
timeline that records:
- 08:30 – Poor sleep; attention
repeatedly drawn to work emails.
- 10:15 – Noticed urge to check
betting app after stressful meeting. Practised attentional shift to a
planned task for 3 minutes; urge reduced.
- 18:20 – Loneliness triggered
thoughts of alcohol. Used urge-surfing exercise and messaged a friend
instead.
- 21:00 – Reflected that cravings
passed without acting on them; confidence rating increased.
Over weeks, the
person would begin to see not only what they felt, but how
attention moved, which cues consistently captured it, and which strategies
successfully redirected it. That creates a personalised map of attentional
vulnerabilities and strengths.
We can't
prevent salient cues from appearing—that's how human perception works—but we
can train people to recognise when attention has been captured and to decide,
deliberately, whether to continue investing that scarce cognitive
resource.
We suspect that
framing the intervention as building attentional control will
align more closely with cognitive science and make the concept more compelling
to clinicians, software developers, and NHS commissioners alike.