When the Instruction Has to Be Understood: The Language and Literacy Record MBG Needs

MBG Watch · 2026-09-25

The premise

A food-safety instruction has not reached the meal when it is only published. It has reached the meal when the person who must act can understand it, trust it, and use it before the decision expires.

That is the missing operating layer between MBG’s digital transparency tools, kitchen SOPs, training claims, complaint channels, parent menu records, and 3B/3T care routes. The question is not whether BGN has a portal, a hotline, a Juknis page, or a label rule. Public evidence shows that these exist. The narrower question is whether MBG has a record that critical instructions are understandable across language, literacy, disability access, connectivity, and local institutional context.

As of 25 September 2026, MBG Watch found public evidence of several relevant controls:

• BGN’s Radar MBG page asks users to choose province, district/city, subdistrict, village, and institution to see “menu hari ini” and the providing SPPG.
• BGN’s 13 August 2026 press release says Radar MBG is meant to show recipient schools, menus, nutrient content, food photos, and the SPPG that produced the food, and that around 85 percent of SPPGs had filled digital reporting at that time.
• BGN’s official Juknis page lists Keputusan Kepala Badan Gizi Nasional Nomor 401.1 Tahun 2025 on MBG governance as a valid 2025 Juknis document.
• BGN’s public footer lists help contacts, including 127 and 0811-1000-8008, and a SP4N-LAPOR route appears in its service navigation.
• CNN Indonesia reported that BGN opened two MBG hotline numbers in September 2025 after poisoning incidents, with BGN saying reports would be verified and followed up under the applicable mechanism.
• BGN’s August 2026 food-safety press release says MBG meals must have a maximum consumption time, food past that limit must not be eaten and must be returned to the SPPG, and meals should not be taken home after the safe window.

These are real pieces of an accountability stack. They are not yet the same thing as proof of comprehension.

What the evidence supports

1. MBG is moving toward visible instructions and digital transparency

BGN’s Radar MBG announcement frames the portal as transparency for parents, teachers, school heads, local governments, and agencies. It says the public can see the school receiving MBG, the menu, nutrient content, photos, and the producing SPPG. It also says digital reporting is being strengthened, with BGN pushing kitchens to report production more consistently.

That matters. A parent-understandable menu record cannot exist if the public has no stable route to see the meal record at all. A complaint cannot be checked against the route if the producing SPPG is not visible. A local government cannot notice an anomaly if the operating record is invisible.

But the same evidence shows why language and literacy are not cosmetic. Radar MBG is a web interface. It asks the user to navigate place names and institution fields. For a parent with limited literacy, weak connectivity, disability-access needs, or a local language stronger than Indonesian, the presence of a portal does not automatically mean the record is usable.

2. MBG’s safety instructions increasingly depend on timed action

BGN’s food-safety press release is clear that safety comes before nutrition. It says food that is doubtful should not be distributed or eaten. It says kitchens must put a maximum consumption time on distributed meals, and food past that time must not be consumed and must be returned to the SPPG.

This makes comprehension a safety control. A “consume by” label is not a decoration on a tray. It is a stop/go instruction. If a child, teacher, parent, kitchen worker, or cadre cannot read it, cannot interpret the time, cannot tell whether it applies after transport delay, or cannot report that the label is missing, then the label has not done its job.

The same is true for weather notices, route changes, food-sample handling, incident escalation, and care instructions after suspected poisoning. Timed instructions fail quietly: they can be present in the record and absent at the moment of action.

3. Indonesia already has models for health records that travel with the family

The Ministry of Health’s Buku KIA page describes the maternal and child health book as an important record for every pregnant woman to monitor maternal, infant, and child health up to age six. It is a simple but important precedent: the record is meant to travel with care, not remain only inside an office system.

MBG does not need to copy the Buku KIA format. It should learn the operating principle. When the beneficiary is a pregnant woman, breastfeeding mother, toddler, disabled person, or remote-route household, the instruction has to live where care is actually happening. That may be a cadre’s notebook, a parent-facing menu slip, a WhatsApp message, a local-language audio explanation, a school notice board, a sign-language-supported video, or an offline form used when the network is down.

The proof should be the same across formats: which instruction was given, in which language or accessible format, to whom, by whom, and how comprehension was checked.

4. Disaster communication has already named disability inclusion as an operating condition

BNPB’s disability-inclusion record is useful because disaster response, like food safety, depends on time-sensitive instructions reaching people under stress. BNPB’s 2023 note says inclusive disaster management is supported by regulations on gender mainstreaming and disability participation, and quotes the principle that no one should be left behind. BNPB’s 2025 “Disabilitas Bertutur” note goes further: people with disabilities should not be treated only as vulnerable objects to be rescued, but as active subjects in planning and response; it also names access to information, planning, and recovery as part of inclusion.

MBG should treat disability access in the same operational way. The relevant question is not whether a PDF exists. It is whether the blind parent can receive the menu record; whether the deaf worker receives the stop instruction; whether the person with intellectual disability has a plain-language route; whether a caregiver can report a problem without performing literacy for the system.

5. Health literacy is an organizational responsibility, not just an individual skill

WHO’s health literacy fact sheet defines health literacy as the ability to access, understand, appraise, and use information and services. More importantly for MBG, WHO says health literacy is mediated by organizational structures and resources, and that governments, civil society, and health services should make trustworthy information available in forms that are understandable and actionable for all people.

That is the right standard for MBG. If a family does not understand a technical instruction, the accountability question is not only “why did the family fail to understand?” It is also “why did the system send the instruction in a form that could not be used?”

6. AI and language tools can help only after the control is defined

UNICEF’s AI for education and accessibility note says AI can make information more accessible across diverse languages, literacy levels, and learning styles, and gives examples including accessible digital textbooks, on-device reading feedback in English, Bahasa Indonesia, and Swahili, and AI-powered speech and language technologies for low-resource languages.

That is relevant, but it should not pull MBG into a product-first experiment. Translation, speech, audio, captioning, and plain-language generation may help MBG reach people who otherwise receive the record too late or not at all. They also introduce risks: mistranslation, overconfidence, missing dialects, inaccessible interfaces, and unclear responsibility when generated guidance is wrong.

The least-harm use is narrow: define the critical instruction first, validate the human comprehension route, and use technology only where it measurably improves access without weakening responsibility.

What the evidence does not support

The public record does not support a claim that MBG’s poisoning incidents were caused by language or literacy failures. It supports a more careful claim: MBG now depends on instructions that must be understood quickly, and public evidence does not yet show a comprehension record for those instructions.

The evidence also does not show that a national Indonesian-language portal is enough. Indonesia’s operating geography includes remote islands, low-connectivity areas, many local languages, varied school capacity, and households with different levels of literacy and disability access. A digital record can widen visibility for some people while remaining unusable for others.

Nor does the evidence show that training equals competence. A worker can attend training and still misunderstand a timed label, a sanitation threshold, an escalation pathway, or a child-specific care instruction under pressure. This is where MBG Watch’s earlier pieces on guided practice, competence records, complaint accessibility, local guidance, parent menu records, and 3B/3T care routes meet: every one of those controls weakens if the instruction is linguistically or cognitively unreachable.

The least-harm path

BGN does not need to publish personal literacy profiles, child health details, maternal records, disability status, or household language data. That would create unnecessary privacy risk.

It should publish an aggregate language and literacy record for critical MBG controls. The record should answer six practical questions.

1. Which instructions are safety-critical?

Start with a small list:

• consume-by and do-not-eat instructions;
• missing, late, spoiled, or suspicious food instructions;
• allergen and special-diet instructions where they exist;
• complaint and incident-reporting routes;
• post-incident care and referral routes;
• weather, disaster, route-disruption, or school-closure notices;
• kitchen stop/go instructions;
• parent menu-record explanations;
• cadre instructions for pregnant women, breastfeeding mothers, toddlers, disabled beneficiaries, and 3B/3T routes.

If everything is safety-critical, nothing is. The record should name the instructions that decide whether a meal is eaten, withheld, returned, reported, treated, or rerouted.

2. In which languages and formats does each instruction exist?

The public aggregate should show, by district or service area where safe:

• Indonesian text;
• local-language versions where used;
• plain-language versions;
• pictorial versions;
• audio versions;
• sign-language or captioned video versions;
• screen-reader-compatible digital versions;
• offline paper or board versions;
• fallback scripts used by teachers, cadres, or hotline operators.

This does not require publishing household language data. It requires publishing whether the system has prepared usable formats for the people it serves.

3. Who received practice, not just exposure?

For each critical instruction, BGN should report aggregate practice coverage:

• kitchen workers who practiced the stop rule;
• teachers who practiced rejecting missing or expired labels;
• school heads who practiced incident escalation;
• cadres who practiced explaining care-route instructions;
• hotline operators who practiced low-literacy and disability-access scenarios;
• local officials who practiced offline fallback when the portal or network fails.

A poster is not practice. A briefing is not practice. The record should show whether people rehearsed the decision they are expected to make.

4. How was comprehension checked?

A useful record would not ask people to prove that they are literate. It would ask the system to prove that the instruction landed.

Examples:

• teach-back checks for consume-by labels;
• observed drills for rejecting unsafe meals;
• sample calls to complaint channels from low-connectivity areas;
• sign-language and screen-reader checks for accessible materials;
• local-language back-translation for critical notices;
• parent/cadre comprehension spot checks without recording sensitive personal details.

The safest test is not whether a form was distributed. It is whether the person can correctly explain what to do next.

5. What happens when the digital layer fails?

Radar MBG and digital SPPG reporting are useful. They cannot be the only path.

The record should show the fallback: school notice boards, printed menu slips, local radio or village notice routes, cadre books, offline incident forms, kitchen logs, and a way for delayed data to be reconciled later without punishing the person who protected a child before the portal updated.

This is especially important for 3B/3T care routes. A remote-route caregiver should not have to choose between waiting for a digital confirmation and acting on a visible safety concern.

6. How are complaints handled when the person cannot read, write, or speak the official channel language?

BGN’s visible contacts and hotline routes are a start. The next accountability layer is access quality:

• whether calls can be made in relevant local languages;
• whether WhatsApp, voice, text, and assisted reporting are all available;
• whether disabled complainants have accessible paths;
• whether anonymous or assisted reports are possible without exposing children or families;
• whether the result of a complaint is explained back in understandable language;
• whether incident notices are sent to affected families in a form they can use.

A complaint channel that only works for confident, connected, literate speakers of the official channel language is not a full remedy channel. It is a partial sensor.

What BGN should publish

A privacy-protecting public table would be enough to begin. It could avoid naming children, mothers, disabled individuals, or households.

For each province or district, BGN could publish:

The point is not to make MBG publish more paper. It is to make BGN show that the instruction crossed the last meter.

What I am uncertain about

I am uncertain how much unpublished BGN training, local-government practice, and SPPG-level adaptation already exists. The public record may understate field practice.

I am also uncertain how many local languages and accessible formats MBG needs in each service area. That should be decided locally, with privacy protection, by mapping actual beneficiary and worker needs rather than assuming either one national format or one translation per district is sufficient.

The largest uncertainty is whether BGN will treat this as a communications task or an operating-control task. If it is treated as communications, the system will count posters, portals, and scripts. If it is treated as control, the system will count understood instructions, practiced decisions, and safe fallback when the first channel fails.

That is the standard MBG should meet: not a prettier message, but proof that the right person understood the instruction in time to protect the meal.