AI Music Therapy for SUD Recovery
T H E S C A L E O F T H E C R I S I S — U S A
All figures from official U.S. government sources: SAMHSA NSDUH 2024 · CDC NCHS 2024 · White House CEA 2025
| 📌 Indicator | 📈 Figure | 📅 Source |
|---|---|---|
| Americans with a past-year SUD (age 12+) | 48.4 million (16.8%) | SAMHSA NSDUH 2024 |
| With Alcohol Use Disorder | 28.1 million (9.7%) | SAMHSA NSDUH 2024 |
| With Drug Use Disorder | 28.4 million (9.8%) | SAMHSA NSDUH 2024 |
| People with SUD who received no treatment | ~80% | SAMHSA NSDUH 2024 |
| Drug overdose deaths in 2024 | 79,384 | CDC NCHS 2024 |
| Decrease in overdose deaths vs 2023 | ↓ 26.2% | CDC NCHS 2024 |
| Annual economic cost of SUD (crime, productivity, healthcare) | $740 billion+ | NIDA / SAMHSA |
| Estimated cost of opioid epidemic alone (2023) | $2.7 trillion (9.7% of GDP) | White House CEA 2025 |
💡 The 80% treatment gap is the single most important number for this research. It is the space — between needing care and receiving it — where AI and music-based interventions, working alongside and in support of FDA-approved medications, are positioned to extend reach, reinforce therapeutic effect, and sustain recovery between clinical contacts.
State-by-State Overdose Picture (CDC 2024 Data)
| 🔴 Highest Burden States | Rate per 100k | 🟢 Largest Declines 2023→2024 | Decline |
|---|---|---|---|
| West Virginia | 81.9 (2023 rate — highest in nation) | West Virginia | ≥ 35% ↓ |
| District of Columbia | 60.7 (2023 rate) | Ohio | ≥ 35% ↓ |
| — | — | Michigan | ≥ 35% ↓ |
| — | — | Louisiana | ≥ 35% ↓ |
| — | — | Virginia | ≥ 35% ↓ |
| — | — | New Hampshire | ≥ 35% ↓ |
| — | — | Wisconsin | ≥ 35% ↓ |
| 🟡 Lowest Burden States | Rate per 100k | ⚠️ States with Increases | Change |
| Nebraska | 9.0 (2023 rate — lowest in nation) | South Dakota | Slight ↑ |
| South Dakota | 11.2 (2023 rate) | Nevada | Slight ↑ |
📍 What drives state variation? Rural vs urban access to MOUD, state-level naloxone laws, opioid prescribing rates, fentanyl trafficking routes, and funding for harm reduction all play documented roles. No single state is “solved” — the 2024 improvements, while significant, still represent nearly 80,000 deaths in a single year.
📊 Full interactive state maps: SAMHSA State Prevalence Maps · CDC Stats of the States
Trends Over Time
DRUG OVERDOSE DEATHS — USA (CDC) ─────────────────────────────────────────────────────────────── 2017 ████████████████████░░░░░░░░░░░░░ ~70,000 2018 ████████████████████░░░░░░░░░░░░░ ~67,000 2019 ████████████████████░░░░░░░░░░░░░ ~70,000 2020 ████████████████████████████░░░░░ ~91,000 ← COVID surge 2021 █████████████████████████████████ ~106,699 ← Peak 2022 ████████████████████████████████░ ~107,500 ← Near-peak 2023 ████████████████████████████░░░░░ ~107,500 → 31.3 per 100k 2024 █████████████████████░░░░░░░░░░░░ 79,384 → 23.1 per 100k ↓26% ─────────────────────────────────────────────────────────────── Fentanyl drove 70%+ of peak deaths. The 2024 decline is real and significant — but 79,384 deaths still exceeds every prior year before 2020. The crisis is not over.
Future Issues & Emerging Threats
| 🚨 Issue | 📋 Detail | 🔭 Outlook |
|---|---|---|
| Polysubstance use | Fentanyl increasingly found mixed with stimulants (meth, cocaine) and benzodiazepines | Complicates treatment; naloxone alone may be insufficient |
| Treatment access gap | 80% of people with SUD receive no treatment — workforce shortages, stigma, cost, geography | Widens without scalable adjunct tools; key driver for this research |
| Equity disparities | Black and Indigenous Americans disproportionately affected; access to MOUD remains unequal | Digital tools risk widening gap if not designed for equity |
| Youth & adolescent SUD | Drug use disorder rates rising among 18–25 age group | Prevention and early intervention urgently needed |
| Mental health comorbidity | ~21.5 million Americans have co-occurring SUD and mental illness | Integrated care models remain under-resourced |
| Xylazine (“tranq”) contamination | Veterinary sedative in the drug supply; no reversal agent exists | Naloxone-resistant overdoses — an emerging harm reduction gap |
| AI & data risks | Poorly validated AI tools entering the recovery space without clinical oversight | Core concern of this research — see RQ5 |
| Post-COVID mental health | Pandemic-driven loneliness, unemployment, trauma amplified SUD risk | Long-tail effects still unfolding through the 2020s |
W H A T T H I S I S A B O U T
People in recovery from Substance Use Disorder face a landscape where professional treatment is essential but often insufficient alone — access is limited, support between sessions is sparse, and the emotional terrain of recovery extends far beyond what medication can reach.
Evidence-based SUD treatment rests on three interlocking pillars:
💊 MEDICATION 🧠 PSYCHOSOCIAL TREATMENT 🤝 PEER & COMMUNITY SUPPORT
───────────────────── ────────────────────────── ────────────────────────
Buprenorphine (OUD) CBT · Motivational Mutual-help groups
Methadone (OUD) Enhancement Therapy Recovery coaching
Naltrexone (OUD · AUD) Contingency Management Community networks
Acamprosate (AUD) 12-step facilitation Peer support workers
FDA-approved · evidence-based Delivered by trained clinicians Lived-experience led
FDA-approved medications reduce craving, prevent withdrawal, and lower overdose risk — but they do not, on their own, rebuild emotional regulation, process trauma, restore identity, or sustain motivation across the months and years of recovery. That is the gap this research addresses.
This study examines how two additional tools can potentiate and extend the effect of those medications — working in the dimensions that pharmacotherapy cannot reach:
🎼 MUSIC-BASED INTERVENTIONS 🤖 ARTIFICIAL INTELLIGENCE
───────────────────────────────── ──────────────────────────────────
Addresses emotional regulation Detects affect states in real time
Processes craving through engagement Predicts lapse risk before it peaks
Rebuilds identity through creativity Delivers support between sessions
Strengthens therapeutic alliance Personalises intensity and timing
Activates reward pathways non-chemically Scales access to underserved regions
Neither literature has adequately spoken to the other — or to how they interact with the medication layer beneath them. This project maps both, surfaces their intersections and contradictions, and proposes an integrative framework that keeps FDA-approved treatment at its foundation.
T H E F O U R - P I L L A R F R A M E W O R K
| Pillar | Core Question | Key Tension | |
|---|---|---|---|
| 🟣 I | Emotional Regulation | Can music and AI-enabled affect tools help people recognise, tolerate, and modulate emotional states without recourse to substance use? | Affect-detection systems claim more certainty than their validation supports |
| 🔴 II | Craving Reduction | What evidence exists that these tools attenuate the urge to use — and when does music induce craving instead? | ⚠️ The craving-cue paradox: music is a conditioned associative stimulus, not a unidirectional tool |
| 🟡 III | Self-Expression | How do songwriting, lyric analysis, improvisation, and AI co-creation support identity reconstruction in recovery? | Generative AI co-creation tools are largely unvalidated in clinical populations |
| 🟢 IV | Personalised Support | How can AI tailor content, timing, and modality to the individual over time — and what are the limits? | Privacy, equity, and over-reliance during crisis are highest stakes here |
⚠️ Critical Design Constraint — The Craving-Cue Paradox
Music can induce craving for alcohol, cannabis, and nicotine by acting as a contextual cue. This is not a footnote — it is a structural constraint running through all four pillars and every phase of this work.
R E S E A R C H Q U E S T I O N S
🔭 OVERARCHING QUESTION — click to expand
In what ways, and with what evidentiary support, can artificial intelligence and music-based interventions function as adjuncts to professional treatment for people with substance use disorder?
| # | Icon | Pillar | Research Question |
|---|---|---|---|
| RQ1 | 🟣 | Emotional Regulation | What is the evidence that music-based interventions and AI-enabled affective tools improve emotional regulation in people with SUD, and through what proposed mechanisms? |
| RQ2 | 🔴 | Craving | What is the evidence that these tools reduce craving — and under what conditions might music instead act as a craving cue? |
| RQ3 | 🟡 | Self-Expression | How do music-based therapeutic techniques and AI co-creation tools support self-expression and identity reconstruction in recovery? |
| RQ4 | 🟢 | Personalisation | How can AI personalise recovery support across time, and what are the limits and risks of doing so? |
| RQ5 | ⚖️ | Ethics (cross-cutting) | What ethical, methodological, and equity considerations should govern the responsible development and study of these tools? |
M E T H O D O L O G I C A L D E S I G N
| 🧩 Dimension | ✅ Decision | 💡 Rationale |
|---|---|---|
| Review Type | PRISMA-ScR Scoping Review | Maps a heterogeneous, emerging, cross-disciplinary field more honestly than a narrow systematic review |
| Synthesis Approach | Separate-literatures synthesis | The AI × Music × SUD intersection is too thin to support a standalone review |
| Population Scope | Substance Use Disorder (SUD) | Full SUD scope; AUD is one part of it where the AI literature is densest, always noted |
| Output | Proposed conceptual framework | The integrative framework — not empirical overlap — is the original contribution |
| Evidence Labelling | Established · Emerging · Speculative | Applied to every key finding; not optional decoration |
🔒 These three scope decisions are locked. Reversing any of them mid-stream is the single most reliable path to timeline collapse.
E V I D E N C E O R I E N T A T I O N
🎵 Music-Based Interventions — State of the Evidence
The most authoritative synthesis is the Cochrane review by Ghetti et al. (2022) — 21 randomised trials (~1,984 participants) reporting moderate-certainty evidence of a medium-sized effect favouring music therapy added to standard care for substance craving. This is the single strongest evidence point in the field.
Earlier reviews (Hohmann et al., 2017; Megranahan & Lynskey, 2018) found beneficial effects on emotional and motivational outcomes — but with substantial inconsistency, heavy reliance on single-session designs, and near-absence of longitudinal trials.
📊 Honest summary: Promising and improving, but heterogeneous and immature.
🤖 AI in SUD Treatment & Recovery — State of the Evidence
The applied AI literature is younger and more fragmented — weighted toward feasibility studies, protocols, and narrative reviews. Active areas include:
- 📱 ML lapse-prediction from EMA and wearable data (Heinz et al., 2025)
- 💬 Conversational agents for around-the-clock support — mixed safety evaluations (Giorgi et al., 2024)
- ⏱️ JITAIs pairing ML predictions with personalised support messages
- 🧬 Affective computing for emotion-aware content adaptation
The most useful framing: Moniz-Lewis et al. (2025) — real promise for equitable, scalable care alongside serious and enumerated ethical concerns.
📊 Honest summary: Real and accelerating activity, genuine promise, but very few rigorously validated tools in clinical deployment.
E T H I C A L F R A M E W O R K
Ethics is not a closing chapter. These are design constraints.
💊 1 · POTENTIATING FRAMING
AI and music-based interventions are studied here as tools that
can enhance and extend the effect of FDA-approved medications
(buprenorphine, naltrexone, methadone, acamprosate) and
psychosocial therapies — addressing dimensions pharmacotherapy
alone cannot reach: emotion, identity, motivation, between-session
support. Nothing here encourages replacing medication with an app.
🎵 2 · THE CRAVING-CUE PARADOX
Music can trigger as well as soothe craving. Any music-based tool
carries a specific, foreseeable risk of harm. Explicit treatment required.
🔒 3 · DATA SENSITIVITY & PRIVACY
Recovery data — location, physiological signals, craving logs — is
among the most sensitive personal data there is. Its disclosure can
carry legal and social consequences for a stigmatised group.
🌍 4 · EQUITY OF ACCESS
Tools dependent on smartphones, wearables, and digital literacy can
widen rather than narrow disparities. Ask: who is served, who excluded?
🚨 5 · RISK OF OVER-RELIANCE
Affect-detection systems present more certainty than they possess.
A conversational agent during a crisis is a genuine safety concern.
Escalation pathways and clinical validation are requirements, not features.
F I V E - P H A S E R E S E A R C H P L A N
┌──────────────────────────────────────────────────────────────────────┐ │ 🟠 PHASE 1 · Protocol & Supervisor Sign-Off ◄ YOU ARE HERE │ ├──────────────────────────────────────────────────────────────────────┤ │ ⚠️ Verify Silverman et al. (2023) — top priority │ │ ✅ Verify Ghetti (2022) and Moniz-Lewis (2025) against primary │ │ 📋 Present protocol skeleton for supervisor sign-off │ │ 🔒 Treat scope decisions as binding once agreed │ └──────────────────────────────────────────────────────────────────────┘ ┌──────────────────────────────────────────────────────────────────────┐ │ 🔵 PHASE 2 · Search Strings, Criteria & Registration │ ├──────────────────────────────────────────────────────────────────────┤ │ 🔍 Build three Boolean strings: music-only / AI-only / AI+music │ │ 📏 Lock inclusion/exclusion criteria │ │ 🛠️ Select appraisal tools (RoB 2, AMSTAR-2, CASP) │ │ 📝 Register on PROSPERO where eligible │ └──────────────────────────────────────────────────────────────────────┘ ┌──────────────────────────────────────────────────────────────────────┐ │ 🟣 PHASE 3 · Search, Screening & Appraisal │ ├──────────────────────────────────────────────────────────────────────┤ │ 🗄️ Run searches across nine databases; deduplicate │ │ 🔎 Two-stage screening: title/abstract → full-text │ │ 📊 Produce PRISMA flow diagram │ │ 👨🏫 Second supervisor checkpoint │ └──────────────────────────────────────────────────────────────────────┘ ┌──────────────────────────────────────────────────────────────────────┐ │ 🟡 PHASE 4 · Synthesis & Framework Construction │ ├──────────────────────────────────────────────────────────────────────┤ │ 🏗️ Synthesise pillar by pillar, then across pillars │ │ 🏷️ Tag every finding: established / emerging / speculative │ │ 🧩 Craving-cue paradox as central finding, not caveat │ │ 🖼️ Construct the proposed conceptual framework │ └──────────────────────────────────────────────────────────────────────┘ ┌──────────────────────────────────────────────────────────────────────┐ │ 🟢 PHASE 5 · Writing, Verification & Submission │ ├──────────────────────────────────────────────────────────────────────┤ │ ✍️ Academic register, prose-led, ethics integrated throughout │ │ 👨🏫 Third supervisor checkpoint: full draft │ │ 🔍 Verification pass: every claim checked against primary source │ │ 🎓 Final submission │ └──────────────────────────────────────────────────────────────────────┘
K E Y O R I E N T I N G R E F E R E N C E S
Starting points only — verify every item against its primary record before citing.
| 👤 Authors | 📅 Year | 📄 Title (abbreviated) | 🏛️ Venue | 🏷️ Pillar | 🔍 Status |
|---|---|---|---|---|---|
| Ghetti et al. | 2022 | Music therapy for people with SUD | Cochrane Database | 🔴 II | ✅ Verify |
| Hohmann et al. | 2017 | Effects of MT and MBIs in SUD treatment | PLOS ONE | 🟣🔴 I·II | ✅ Verify |
| Megranahan & Lynskey | 2018 | Do creative arts therapies reduce misuse? | Arts in Psychotherapy | 🟡 III | ✅ Verify |
| Silverman et al. | 2023 | Music as a cue for substance craving | TBC | 🔴 II | ⚠️ UNVERIFIED |
| Moniz-Lewis et al. | 2025 | AI in alcohol research and treatment | Alcohol Research: CR | 🟢⚖️ IV·V | ✅ Verify |
| Giorgi et al. | 2024 | Generative AI responses to drug questions | Psychiatry Research | 🟢 IV | ✅ Verify |
| Heinz et al. | 2025 | EMA + deep learning for opioid prediction | JSUAT | 🔴🟢 II·IV | ✅ Verify |
| De Freitas & Cohen | 2024 | Health risks of generative AI wellness apps | Nature Medicine | 🟢⚖️ IV·V | ✅ Verify |
L A N G U A G E & S T A N D A R D S
| 🏷️ Standard | 📌 Rule |
|---|---|
| 🧑 Person-First Language | “People with a substance use disorder” — never “addicts” or “users” |
| 🎵 Terminological Precision | MT ≠ MBI ≠ consumer listening — maintained every time it matters |
| 📊 Evidence Calibration | Established · Emerging · Speculative — labelled in every key claim |
| 🔍 Verify Before Asserting | No claim reproduced from memory — all checked against primary sources |
| ⚖️ Ethics Integrated | Caveats sit next to the claims they qualify, not in a closing section |
C I T I N G T H I S W O R K
A CITATION.cff file is included so GitHub provides a “Cite this repository” option.
Until the work is verified and complete, treat any citation as referring to an in-progress draft.
L I C E N S I N G
No license is currently applied — all rights reserved by default.
A Creative Commons option may be added pending institutional IP guidance.
Get in touch
Bernard Consulting
Want to work with Bernard on your digital projects? Send him a message to let him know what your challenges are! You can reach out to Bernard via his Linkedin profile. https://www.linkedin.com/in/bernard-g/
Email
bernard@vgoshinfo.com
Phone:
+1 (954)-804-4785
Address:
Vgosh Info LLC, 111 NE, 1st Street,
8th Floor, 88510, Miami, FL 33142
8th Floor, 88510, Miami, FL 33142