Draft — Digital Humanism Fellowship 2026–2027

MOTIVATION LETTER

The Digital Humanism Fellowship asks how digital technologies can be built to serve human dignity, democratic values, and social well-being. My research answers that question from a position most applicants do not occupy: a Nigerian pharmacist and computational modeler who has spent four years building quantitative models of addiction, drug resistance, and brain-circuit dynamics, and who is now enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute in Potsdam. The CCT model, my primary research line, is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It couples dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast into a single ODE system, calibrated with Bayesian MCMC against a literature screen of 1,847 records. All five pre-registered hypotheses were confirmed, with posterior super-additivity of 13 to 22 percentage points across model versions. This is a computational foundation for designing digital mental health interventions that do not replicate the biases of the clinics they replace.

The fellowship's emphasis on interdisciplinary research matches my training. I hold a B.Pharm from the University of Ibadan, work daily in Python, PyMC, and topological data analysis, and have published negative results when they were the honest outcome. My hERG cardiotoxicity study, pre-registered and powered, showed that bipartite persistent homology does not beat a plain descriptor baseline for predicting toxicity from protein-ligand interface geometry. That finding settled a comparison the literature had never actually run. I bring the same discipline to the societal questions this fellowship prioritizes: algorithmic bias in mental health apps, transparency in clinical decision support, and the ethics of deploying Western-trained models in African healthcare systems.

The Nigeria angle is a design constraint. Digital health tools built for Lagos or Ibadan must function where connectivity is intermittent, where regulatory oversight is thin, and where the default assumption of a patient-facing algorithm is distrust. My work on the TOPOLOGIX pipeline, which predicts drug-resistance mutations from sequence alone using ESM-2 protein language model embeddings, covers 100 percent of mutations in the Platinum benchmark versus roughly 18 percent for structure-limited tools. That coverage gap matters clinically in settings where structural data does not exist for local pathogen strains. The same logic applies to mental health: a model calibrated on European populations cannot be assumed to generalize to Nigerian patients without explicit validation. The CCT framework, built from first principles in pharmacology, is a step toward models that are transparent enough to audit and mechanistic enough to explain.

IWM's residency structure, with its seminars and scholarly community, is the environment I need to translate these technical results into a research agenda that engages philosophy, ethics, and policy. My collaborators include Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am not asking for a platform to present finished work. I am asking for a community to test whether my models can be made legible to scholars outside computational neuroscience, and whether the ethical questions embedded in my methods can be made rigorous enough to survive their scrutiny.

RESEARCH STATEMENT

The CCT model, Conjunctive Consolidation Threshold, addresses a specific clinical failure: relapse in addiction. Current digital mental health tools, from chatbot interventions to smartphone-based cognitive behavioral therapy, operate without a mechanistic account of how reward memories are consolidated. They treat addiction as a behavioral pattern to be interrupted, not as a neurobiological process with a definable threshold for memory encoding. My research asks whether that threshold can be modeled computationally, and whether the resulting model can inform the design of digital interventions that are both effective and ethically defensible.

The model itself is a system of coupled ordinary differential equations with three axes. The dopaminergic axis encodes reward prediction error, the NMDAR axis encodes synaptic plasticity via long-term potentiation, and the affective contrast axis encodes the emotional valence that gates memory strength. I calibrated the system using Bayesian MCMC with PyMC's DEMetropolisZ sampler, fitting 14 free parameters against priors elicited from a systematic screen of 1,847 published records. The model was pre-registered with five hypotheses before calibration. All five were confirmed. The posterior distribution shows super-additivity of 13 to 22 percentage points across model versions, meaning the combined effect of the three axes exceeds the sum of their individual contributions. That result has a direct design implication: interventions that target only one axis, such as dopaminergic blockade alone, are mathematically insufficient. The threshold is conjunctive.

The fellowship's selection criteria emphasize investigation of societal and ethical implications, not merely application of digital tools. My research speaks to that criterion in two ways. First, the CCT model is a transparency instrument. Unlike black-box machine learning classifiers, which dominate current mental health app design, the CCT model exposes its assumptions in the form of differential equations and prior distributions. A clinician can inspect why the model predicts a given intervention will fail. That inspectability is an ethical property, not just a technical one. Second, the model is being developed with African deployment contexts in mind. Nigeria has fewer than 300 psychiatrists for a population exceeding 200 million. Digital tools are not optional there; they are the only scalable option. But a tool that embeds unexamined Western assumptions about reward, memory, and motivation will fail clinically and ethically. The CCT model, built from pharmacological first principles, is a candidate foundation for tools that can be re-calibrated to local populations.

My broader methodological record supports the feasibility of this agenda. I have built neurocascade, a receptor-to-behavior simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics, with 62 of 62 tests passing. I have developed TOPOLOGIX, which predicts drug-resistance mutations from sequence alone with an AUROC of 0.804 on the Platinum benchmark. I have also published negative results, including the hERG cardiotoxicity study where topological features failed to beat a descriptor baseline, and the ergofluids project where the first real-data gate did not meet its pre-registered criterion. I report failures directly. That practice is essential for a field where publication bias distorts the evidence base for digital mental health.

During the three-month residency at IWM, I will produce three deliverables. First, a working paper translating the CCT model's mathematical structure into a set of design principles for digital addiction interventions, written for an interdisciplinary audience. Second, a public repository containing the full model code, calibration data, and pre-registration documents, so that other researchers can audit and extend the work. Third, a seminar presentation that maps the model's ethical implications, specifically the question of who decides what counts as a reward memory worth preventing. That question is clinical, technical, and philosophical. IWM is the right place to ask it.

SHORT ESSAY: INTERDISCIPLINARY APPROACH

My training spans pharmacy, computational neuroscience, and software engineering. That combination is unusual, and it is the reason I can engage the Digital Humanism agenda substantively rather than rhetorically. I have dispensed medications in Nigerian community pharmacies, where I watched patients cycle through addiction treatment with no digital support whatsoever. I have also built production-grade data pipelines and self-hosted LLM infrastructure, which means I know how the technology works from the inside. The gap between those two experiences is the gap this fellowship exists to address.

The CCT model is inherently interdisciplinary. It requires pharmacology to specify the receptor systems, dynamical systems theory to specify the equations, Bayesian statistics to calibrate the parameters, and ethics to ask whether the model should be used to predict relapse risk in individual patients. I have published in all four registers: a co-authored paper in Alcohol, three sole-authored preprints under review at peer-reviewed journals, and open-source code repositories with full documentation. My collaborators include Kent Berridge, whose work on incentive salience defined the field, and Samuel Gershman, who endorsed my arXiv submissions. I am not an outsider asking to be let in. I am a practitioner asking for a community that can hold my work to standards I cannot apply alone.

The fellowship's emphasis on combining computer science with humanities and social science perspectives is a practical necessity for me. My models produce numbers. They do not produce meaning. The question of whether a relapse prediction should be shared with a patient, a family member, or a parole officer is not a statistical question. It is a legal, ethical, and cultural question. I need philosophers and legal scholars to help me ask it properly, and IWM's seminar structure is designed for exactly that exchange.

SHORT ESSAY: AFRICA AND DIGITAL HUMANISM

Digital humanism cannot be a global movement if it ignores Africa. The continent has the youngest population in the world, the fastest-growing mobile phone adoption, and the most severe healthcare workforce shortages. Digital health tools are being deployed across Nigeria, Kenya, and South Africa at scale, often built by Western companies with Western training data and Western assumptions about user behavior. The result is a form of technological colonialism: algorithms that work in Boston or Berlin are exported to Lagos without validation, without local calibration, and without accountability.

My research is a counterexample. The CCT model is built from pharmacological mechanisms that are universal, not from population-specific behavioral data. The equations describe how dopamine, NMDA receptors, and affective contrast interact to consolidate reward memories. Those mechanisms exist in every human brain. But the parameters, the prior distributions, and the clinical thresholds must be calibrated locally. That is the design principle I bring to Digital Humanism: universal mechanism, local calibration. It is the same principle that drives my TOPOLOGIX work, where sequence-based prediction covers 100 percent of mutations in the Platinum benchmark because it does not depend on structural data that rarely exists for African pathogen strains.

I am also honest about the limits of my position. I am a Nigerian researcher, but I am currently enrolled in a German master's program and collaborating with American academics. I do not speak for Africa. I speak from a specific vantage point: a pharmacist who has worked in Nigerian clinical settings and a modeler who knows what the technology can and cannot do. The fellowship would allow me to test whether my design principles survive contact with scholars who study technology policy, democratic governance, and social well-being from perspectives I have not yet encountered. That test is the point.

CHECKLIST

- [ ] Confirm eligibility: must be enrolled in a doctoral program OR have obtained a PhD no earlier than 21 September 2022. Applicant is enrolled in M.Sc. Digital Health at HPI/Potsdam, not a doctoral program, and does not hold a PhD. Verify whether the M.Sc. enrollment satisfies the doctoral enrollment requirement or whether an exception applies.
- [ ] Verify the application deadline and any country-specific eligibility restrictions on the official IWM Digital Humanism Fellowship page, not the reigngigs.com aggregator URL.
- [ ] Obtain two letters of recommendation. Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar are named as collaborators/endorsers; confirm which two are willing to write and submit letters by the deadline.
- [ ] Prepare a CV in the format required by IWM, including ORCID, GitHub, personal site, publications, and the three preprints currently under review.
- [ ] Confirm the exact word limits for the motivation letter, research statement, and short essays on the official application portal; adjust the drafts above if limits differ from the 300-500 word and 200-350 word ranges used here.
- [ ] Verify the status of the three sole-authored preprints (IART, PNPBP, NBR) and the co-authored Alcohol paper; update the research statement with acceptance or rejection outcomes if any have changed.
- [ ] Confirm the start date and duration of the fellowship residency at IWM in Vienna, and check whether the M.Sc. program at HPI/Potsdam permits a three-month leave or remote study period.
- [ ] Prepare a project timeline for the three-month residency, mapping the CCT working paper, public repository release, and seminar presentation to specific weeks.
- [ ] Check whether IWM requires a separate project proposal form or budget outline in addition to the motivation letter and research statement; prepare if required.

EDITOR NOTES

- Eligibility risk is the single largest issue. The fellowship specifies doctoral or postdoctoral status. The applicant is enrolled in a master's program and holds a B.Pharm. The letter does not resolve this. The applicant must either confirm that M.Sc. enrollment qualifies, or find a doctoral program to enroll in before applying, or accept that this application may be rejected on eligibility grounds regardless of quality.
- The CCT model is presented as the primary research line, consistent with the recommended framing angle. The hERG cardiotoxicity study is mentioned as a negative result to demonstrate rigor. Verify that the hERG study is indeed published or posted as a preprint with a DOI, and that the AUROC figures (0.8426 vs 0.8782) are accurate, before submitting.
- The letter claims the CCT model has "all five pre-registered hypotheses confirmed" and cites "posterior super-additivity 13-22pp." These are strong quantitative claims. The applicant must verify these numbers against the actual preprints and be prepared to provide the OSF/Zenodo links in the application portal.
- The Africa framing is present but could be sharpened with one concrete example of a Nigerian digital health deployment that failed due to lack of local calibration. The applicant should insert a specific named case if one exists in their professional experience, or soften the claim to avoid overgeneralization.
- The short essay on interdisciplinary approach mentions "self-hosted LLM infrastructure" and "production-grade data pipelines." These are technical skills, not research contributions. If the selection committee values academic output over engineering capability, these claims may read as padding. The applicant should consider cutting or compressing them in favor of more detail on the CCT model's ethical implications.
- The research statement promises three deliverables for the three-month residency. Confirm that IWM expects concrete deliverables of this kind, or adjust the promises to match the fellowship's actual structure (seminars, working papers, public lectures). Overpromising is a common rejection reason.
- The motivation letter opens with the fellowship's mission rather than the applicant's name or biography. This is intentional per the formatting rules, but it means the letter does not introduce the applicant until the second sentence. Confirm this stylistic choice is acceptable to the applicant before submission.

MOTIVATION LETTER

The Digital Humanism Fellowship asks how digital technologies can be built to serve human dignity, democratic values, and social well-being. My research answers that question from a position most applicants do not occupy: a Nigerian pharmacist and computational modeler who has spent four years building quantitative models of addiction, drug resistance, and brain-circuit dynamics, and who is now enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute in Potsdam. The CCT model, my primary research line, is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It couples dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast into a single ODE system, calibrated with Bayesian MCMC against a literature screen of 1,847 records. All five pre-registered hypotheses were confirmed, with posterior super-additivity of 13 to 22 percentage points across model versions. This is a computational foundation for designing digital mental health interventions that do not replicate the biases of the clinics they replace.

The fellowship's emphasis on interdisciplinary research matches my training. I hold a B.Pharm from the University of Ibadan, work daily in Python, PyMC, and topological data analysis, and have published negative results when they were the honest outcome. My hERG cardiotoxicity study, pre-registered and powered, showed that bipartite persistent homology does not beat a plain descriptor baseline for predicting toxicity from protein-ligand interface geometry. That finding settled a comparison the literature had never actually run. I bring the same discipline to the societal questions this fellowship prioritizes: algorithmic bias in mental health apps, transparency in clinical decision support, and the ethics of deploying Western-trained models in African healthcare systems.

The Nigeria angle is a design constraint. Digital health tools built for Lagos or Ibadan must function where connectivity is intermittent, where regulatory oversight is thin, and where the default assumption of a patient-facing algorithm is distrust. My work on the TOPOLOGIX pipeline, which predicts drug-resistance mutations from sequence alone using ESM-2 protein language model embeddings, covers 100 percent of mutations in the Platinum benchmark versus roughly 18 percent for structure-limited tools. That coverage gap matters clinically in settings where structural data does not exist for local pathogen strains. The same logic applies to mental health: a model calibrated on European populations cannot be assumed to generalize to Nigerian patients without explicit validation. The CCT framework, built from first principles in pharmacology, is a step toward models that are transparent enough to audit and mechanistic enough to explain.

IWM's residency structure, with its seminars and scholarly community, is the environment I need to translate these technical results into a research agenda that engages philosophy, ethics, and policy. My collaborators include Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am not asking for a platform to present finished work. I am asking for a community to test whether my models can be made legible to scholars outside computational neuroscience, and whether the ethical questions embedded in my methods can be made rigorous enough to survive their scrutiny.

RESEARCH STATEMENT

The CCT model, Conjunctive Consolidation Threshold, addresses a specific clinical failure: relapse in addiction. Current digital mental health tools, from chatbot interventions to smartphone-based cognitive behavioral therapy, operate without a mechanistic account of how reward memories are consolidated. They treat addiction as a behavioral pattern to be interrupted, not as a neurobiological process with a definable threshold for memory encoding. My research asks whether that threshold can be modeled computationally, and whether the resulting model can inform the design of digital interventions that are both effective and ethically defensible.

The model itself is a system of coupled ordinary differential equations with three axes. The dopaminergic axis encodes reward prediction error, the NMDAR axis encodes synaptic plasticity via long-term potentiation, and the affective contrast axis encodes the emotional valence that gates memory strength. I calibrated the system using Bayesian MCMC with PyMC's DEMetropolisZ sampler, fitting 14 free parameters against priors elicited from a systematic screen of 1,847 published records. The model was pre-registered with five hypotheses before calibration. All five were confirmed. The posterior distribution shows super-additivity of 13 to 22 percentage points across model versions, meaning the combined effect of the three axes exceeds the sum of their individual contributions. That result has a direct design implication: interventions that target only one axis, such as dopaminergic blockade alone, are mathematically insufficient. The threshold is conjunctive.

The fellowship's selection criteria emphasize investigation of societal and ethical implications, not merely application of digital tools. My research speaks to that criterion in two ways. First, the CCT model is a transparency instrument. Unlike black-box machine learning classifiers, which dominate current mental health app design, the CCT model exposes its assumptions in the form of differential equations and prior distributions. A clinician can inspect why the model predicts a given intervention will fail. That inspectability is an ethical property, not just a technical one. Second, the model is being developed with African deployment contexts in mind. Nigeria has fewer than 300 psychiatrists for a population exceeding 200 million. Digital tools are not optional there; they are the only scalable option. But a tool that embeds unexamined Western assumptions about reward, memory, and motivation will fail clinically and ethically. The CCT model, built from pharmacological first principles, is a candidate foundation for tools that can be re-calibrated to local populations.

My broader methodological record supports the feasibility of this agenda. I have built neurocascade, a receptor-to-behavior simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics, with 62 of 62 tests passing. I have developed TOPOLOGIX, which predicts drug-resistance mutations from sequence alone with an AUROC of 0.804 on the Platinum benchmark. I have also published negative results, including the hERG cardiotoxicity study where topological features failed to beat a descriptor baseline, and the ergofluids project where the first real-data gate did not meet its pre-registered criterion. I report failures directly. That practice is essential for a field where publication bias distorts the evidence base for digital mental health.

During the three-month residency at IWM, I will produce three deliverables. First, a working paper translating the CCT model's mathematical structure into a set of design principles for digital addiction interventions, written for an interdisciplinary audience. Second, a public repository containing the full model code, calibration data, and pre-registration documents, so that other researchers can audit and extend the work. Third, a seminar presentation that maps the model's ethical implications, specifically the question of who decides what counts as a reward memory worth preventing. That question is clinical, technical, and philosophical. IWM is the right place to ask it.

SHORT ESSAY: INTERDISCIPLINARY APPROACH

My training spans pharmacy, computational neuroscience, and software engineering. That combination is unusual, and it is the reason I can engage the Digital Humanism agenda substantively rather than rhetorically. I have dispensed medications in Nigerian community pharmacies, where I watched patients cycle through addiction treatment with no digital support whatsoever. I have also built production-grade data pipelines and self-hosted LLM infrastructure, which means I know how the technology works from the inside. The gap between those two experiences is the gap this fellowship exists to address.

The CCT model is inherently interdisciplinary. It requires pharmacology to specify the receptor systems, dynamical systems theory to specify the equations, Bayesian statistics to calibrate the parameters, and ethics to ask whether the model should be used to predict relapse risk in individual patients. I have published in all four registers: a co-authored paper in Alcohol, three sole-authored preprints under review at peer-reviewed journals, and open-source code repositories with full documentation. My collaborators include Kent Berridge, whose work on incentive salience defined the field, and Samuel Gershman, who endorsed my arXiv submissions. I am not an outsider asking to be let in. I am a practitioner asking for a community that can hold my work to standards I cannot apply alone.

The fellowship's emphasis on combining computer science with humanities and social science perspectives is a practical necessity for me. My models produce numbers. They do not produce meaning. The question of whether a relapse prediction should be shared with a patient, a family member, or a parole officer is not a statistical question. It is a legal, ethical, and cultural question. I need philosophers and legal scholars to help me ask it properly, and IWM's seminar structure is designed for exactly that exchange.

SHORT ESSAY: AFRICA AND DIGITAL HUMANISM

Digital humanism cannot be a global movement if it ignores Africa. The continent has the youngest population in the world, the fastest-growing mobile phone adoption, and the most severe healthcare workforce shortages. Digital health tools are being deployed across Nigeria, Kenya, and South Africa at scale, often built by Western companies with Western training data and Western assumptions about user behavior. The result is a form of technological colonialism: algorithms that work in Boston or Berlin are exported to Lagos without validation, without local calibration, and without accountability.

My research is a counterexample. The CCT model is built from pharmacological mechanisms that are universal, not from population-specific behavioral data. The equations describe how dopamine, NMDA receptors, and affective contrast interact to consolidate reward memories. Those mechanisms exist in every human brain. But the parameters, the prior distributions, and the clinical thresholds must be calibrated locally. That is the design principle I bring to Digital Humanism: universal mechanism, local calibration. It is the same principle that drives my TOPOLOGIX work, where sequence-based prediction covers 100 percent of mutations in the Platinum benchmark because it does not depend on structural data that rarely exists for African pathogen strains.

I am also honest about the limits of my position. I am a Nigerian researcher, but I am currently enrolled in a German master's program and collaborating with American academics. I do not speak for Africa. I speak from a specific vantage point: a pharmacist who has worked in Nigerian clinical settings and a modeler who knows what the technology can and cannot do. The fellowship would allow me to test whether my design principles survive contact with scholars who study technology policy, democratic governance, and social well-being from perspectives I have not yet encountered. That test is the point.

CHECKLIST

- [ ] Confirm eligibility: must be enrolled in a doctoral program OR have obtained a PhD no earlier than 21 September 2022. Applicant is enrolled in M.Sc. Digital Health at HPI/Potsdam, not a doctoral program, and does not hold a PhD. Verify whether the M.Sc. enrollment satisfies the doctoral enrollment requirement or whether an exception applies.
- [ ] Verify the application deadline and any country-specific eligibility restrictions on the official IWM Digital Humanism Fellowship page, not the reigngigs.com aggregator URL.
- [ ] Obtain two letters of recommendation. Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar are named as collaborators/endorsers; confirm which two are willing to write and submit letters by the deadline.
- [ ] Prepare a CV in the format required by IWM, including ORCID, GitHub, personal site, publications, and the three preprints currently under review.
- [ ] Confirm the exact word limits for the motivation letter, research statement, and short essays on the official application portal; adjust the drafts above if limits differ from the 300-500 word and 200-350 word ranges used here.
- [ ] Verify the status of the three sole-authored preprints (IART, PNPBP, NBR) and the co-authored Alcohol paper; update the research statement with acceptance or rejection outcomes if any have changed.
- [ ] Confirm the start date and duration of the fellowship residency at IWM in Vienna, and check whether the M.Sc. program at HPI/Potsdam permits a three-month leave or remote study period.
- [ ] Prepare a project timeline for the three-month residency, mapping the CCT working paper, public repository release, and seminar presentation to specific weeks.
- [ ] Check whether IWM requires a separate project proposal form or budget outline in addition to the motivation letter and research statement; prepare if required.

EDITOR NOTES

- Eligibility risk is the single largest issue. The fellowship specifies doctoral or postdoctoral status. The applicant is enrolled in a master's program and holds a B.Pharm. The letter does not resolve this. The applicant must either confirm that M.Sc. enrollment qualifies, or find a doctoral program to enroll in before applying, or accept that this application may be rejected on eligibility grounds regardless of quality.
- The CCT model is presented as the primary research line, consistent with the recommended framing angle. The hERG cardiotoxicity study is mentioned as a negative result to demonstrate rigor. Verify that the hERG study is indeed published or posted as a preprint with a DOI, and that the AUROC figures (0.8426 vs 0.8782) are accurate, before submitting.
- The letter claims the CCT model has "all five pre-registered hypotheses confirmed" and cites "posterior super-additivity 13-22pp." These are strong quantitative claims. The applicant must verify these numbers against the actual preprints and be prepared to provide the OSF/Zenodo links in the application portal.
- The Africa framing is present but could be sharpened with one concrete example of a Nigerian digital health deployment that failed due to lack of local calibration. The applicant should insert a specific named case if one exists in their professional experience, or soften the claim to avoid overgeneralization.
- The short essay on interdisciplinary approach mentions "self-hosted LLM infrastructure" and "production-grade data pipelines." These are technical skills, not research contributions. If the selection committee values academic output over engineering capability, these claims may read as padding. The applicant should consider cutting or compressing them in favor of more detail on the CCT model's ethical implications.
- The research statement promises three deliverables for the three-month residency. Confirm that IWM expects concrete deliverables of this kind, or adjust the promises to match the fellowship's actual structure (seminars, working papers, public lectures). Overpromising is a common rejection reason.
- The motivation letter opens with the fellowship's mission rather than the applicant's name or biography. This is intentional per the formatting rules, but it means the letter does not introduce the applicant until the second sentence. Confirm this stylistic choice is acceptable to the applicant before submission.

AI Draft — Digital Humanism Fellowship 2026–2027