# Voices of the week

**The weekly conversation · 21–27 October 2026**

Ideas worth spending time with.

**Who gets to shape the answer?**

![Energetic emerald and azure brushstrokes converge around dark green marks and small ivory openings.](/assets/publications/voices/2026-10-27/hero.png)

*Original artwork: The Shape of Agency. Chiappe × OpenAI.*

A useful AI system is not simply one that produces an answer. This week's voices ask who can question that answer, what people learn from using it, who pays to keep the service working and whose needs shape its direction.

**In this issue**

## Learning that survives without the tool

![A dense blue-green painted field faces an independent emerald stroke across an ivory gap.](/assets/publications/voices/2026-10-27/schwartz-hakala.png)

*Original artwork: What Remains. Chiappe × OpenAI.*

**Beth M. Schwartz and Christopher Hakala**  
APA senior director and cognitive psychologist; Springfield College learning-science researcher  
**Education · Expert opinion · 22 October 2026**

**The argument:** Established learning science can guide AI use before tool-specific evidence matures. Protect the mental work required for the intended learning.

**What is new:** The useful test is not merely better assisted work. Can students demonstrate and transfer understanding without AI afterwards?

**Why it matters:** A polished assignment and durable learning are different outcomes.

**LecXie’s practical interpretation:** Try an unaided attempt, AI feedback and an explanation of the student's choices. Then assess a new task without assistance.

**Evidence & limits:** Expert interpretation, not a new AI trial or quantified benefit. The authors call for small studies across disciplines and learner populations. Our suggested sequence is a testable design, not a proven intervention.

[Original contribution](https://www.insidehighered.com/opinion/views/2026/10/22/science-learning-should-guide-ai-use-decisions-opinion) · [Schwartz at APA](https://www.apa.org/applied-psychology) · [Hakala at Springfield College](https://springfield.edu/directory/chris-hakala)

## Fund the service, not just the demonstration

![Thick dark green and ochre layers support a lighter burst of gold and blue paint.](/assets/publications/voices/2026-10-27/barenblat.png)

*Original artwork: The Work Beneath. Chiappe × OpenAI.*

**Kevin Barenblat**  
Co-founder and president, Fast Forward  
**Social purpose · Sector commentary · 22 October 2026**

**The argument:** Nonprofits need sustained infrastructure funding to turn promising AI tools into dependable services.

**What is new:** Barenblat shifts attention from inexpensive prototypes to the less visible costs of staffing, maintenance and responsible operation.

**Why it matters:** A successful demonstration does not establish that an organisation can support the people relying on it.

**LecXie’s practical interpretation:** Before financing a pilot, identify who will maintain it, evaluate errors and meet recurring costs after the initial grant.

**Evidence & limits:** Commentary from an organisation supporting technology nonprofits. Its underlying report was released in September, not this coverage week. The report uses a self-reported sample of AI-focused nonprofits rather than a representative comparison of the whole sector. Google.org supported the research. This is not proof that AI is the best use of every charity's funds.

[Original contribution](https://candid.org/blogs/ai-for-humanity-2026-report/) · [Author profile](https://candid.org/authors/kevin-barenblat/) · [Report methods and findings](https://www.ffwd.org/blog/ai-for-nonprofits-2026-report-findings)

## Local value matters more than adoption totals

![Azure currents meet a concentrated emerald centre within a warm terracotta and ochre field.](/assets/publications/voices/2026-10-27/ohemeng-boamah.png)

*Original artwork: Value Held Here. Chiappe × OpenAI.*

**Anthony Ohemeng-Boamah**  
International-development writer  
**African development · Policy opinion · 23 October 2026**

**The argument:** Judge African AI by locally solved problems and locally retained value, not adoption totals alone.

**What is new:** Ohemeng-Boamah uses mobile money as a design analogy: access and interoperability can matter more than sophistication.

**Why it matters:** More use of imported platforms need not mean more local agency.

**LecXie’s practical interpretation:** Before procurement, name the local need, accessible delivery route and interoperability requirements. Ask who retains the resulting capability and value.

**Evidence & limits:** A policy argument, not evidence that payment-system successes have already been replicated in AI. The UN's Guinea roster lists the author as its UNDP representative. The contribution should not be read as an institutional endorsement of this selection or as proof of UNDP policy.

[Original contribution](https://mg.co.za/thought-leader/2026-10-22-africa-s-ai-moment-must-be-more-than-imported-intelligence/) · [Official country-team listing](https://unsdg.un.org/un-in-action/guinea)

## Keep judgement in the public library

![Deep green strokes interrupt layered blue and ivory paint, punctuated by small red marks.](/assets/publications/voices/2026-10-27/ferreiro.png)

*Original artwork: Between Answer and Understanding. Chiappe × OpenAI.*

**Carmen Ferreiro**  
Director, A Coruña Provincial Library  
**Public services · Practitioner interview · 23 October 2026**

**The argument:** Libraries need professional judgement, not just AI-generated answers.

**What is new:** Ferreiro describes emerging uses while warning that polished answers can conceal unreliable information.

**Why it matters:** Public access to knowledge depends on skilled human guidance.

**LecXie’s practical interpretation:** Preserve librarian-led verification when introducing AI-assisted information services.

**Evidence & limits:** A reported practitioner interview, not an evaluated deployment. Ferreiro speaks from the library profession. English translated summary.

[Original Spanish interview report](https://cadenaser.com/galicia/2026/10/23/las-bibliotecas-reivindican-su-papel-mas-alla-de-los-libros-hay-mucho-trabajo-detras-que-no-se-ve-radio-coruna/) · [Institutional professional programme](https://www.dacoruna.gal/biblioteca/bpdc-para-profesionais/)

## Who chooses the direction?

![A broad blue current branches into uneven green and ochre sweeps across textured ivory canvas.](/assets/publications/voices/2026-10-27/cavello.png)

*Original artwork: Direction Is a Choice. Chiappe × OpenAI.*

**B Cavello**  
Director of emerging technologies, Aspen Digital  
**Technology governance · Policy argument · 27 October 2026**

**The argument:** Changing AI's speed is insufficient if its destination is set without the people affected.

**What is new:** Cavello moves the debate from whether development should accelerate or slow down to who can influence its priorities.

**Why it matters:** Greater capability does not automatically improve worker safety, privacy or room to learn.

**LecXie’s practical interpretation:** Before a pilot, ask affected staff what should improve and which uses concern them. Record how those answers change the trial.

**Evidence & limits:** A policy argument drawing on community engagement and earlier attitude research, not an evaluated governance intervention. Cavello advocates participatory technology governance through Aspen Digital. Consultation is not itself proof that people have meaningful influence.

[Original contribution](https://www.aspendigital.org/blog/its-not-the-speed-its-the-direction/) · [Official profile](https://www.aspendigital.org/person/b-cavello/)

### The thread between them

These contributions ask different questions about what surrounds an AI answer: independent understanding, sustainable funding, local agency, professional judgement and influence over development. Together they suggest a practical discipline. Define the human outcome first. Then establish who can challenge the system and who remains responsible when it fails.

That is our synthesis, not a shared manifesto or an established causal finding. The voices here offer arguments and practitioner perspectives rather than a set of measured interventions. Their ideas deserve testing against local circumstances and against non-AI alternatives.

### How we choose

Original contributions published during 21–27 October 2026 were considered for relevance, distinctness and a useful implication for work, learning or public life. Dates refer to the contributions, not necessarily the research they discuss. Official profiles or institutional identity sources are linked where personal profiles could not be verified. Inclusion is not endorsement, collaboration or a ranking of the people involved.
