What is Adaptive Administration?

The book *Plurality*'s vision of using technologies such as text-generating AI to let the fixed forms and rules of government offices and large companies adapt to each user's own words and circumstances. The rules that ensure fairness stay in place, while diverse people can use the system as they are.

Have you ever been stuck at a government counter, unsure how to fill in a form? Unless you write the expected words in the expected boxes, nothing moves forward. It is harder still for people who are not fluent in the language, or whose situation does not fit the template. Adaptive Administration is the vision, in the book ⿻ Plurality, of using technology to lower this wall of forms.

In short

It is the vision of letting the systems of government offices and large companies, which process things through fixed forms and rules, adapt to each user's own words and circumstances [1][2].

As the tool for this, the book names AI that generates text and more (in the book's terms, generative foundation models, or GFMs). The aim is to keep the rules that ensure fairness while letting diverse people use the system as they are [1]. "Adaptive" means changing shape to suit the person or situation. The book's chapter summary puts it this way: "'Adaptive administration' builds translators across administrative and legal systems, allowing people to live in a wider diversity of ways while still collaborating with each other under the rule of law" [2].

For example

An Indian farmer and a voice AI. At the World Economic Forum in Davos, Switzerland, Microsoft CEO Satya Nadella gave a demonstration. A farmer in rural India speaks into a feature phone in a local language. A large language model (AI that understands and produces text) "understood the voice, translated from the local language to the national language in which the relevant forms were available, helped navigate what needed to be filled out, and returned guidance via voice to the farmer" [3].

The book writes that this built on years of collaboration, and holds "the promise of helping preserve and strengthen the cultural diversity of India by ensuring those who speak less prominent languages and live far from cities are still able to access the public services they need". Since then, a government chatbot to support applications to farmer financial support programs and multilingual guidance on a messaging app have been launched [3].

Chatbots that guide people to benefits. As earlier examples, the book names chatbots from Finland's social insurance institution, Germany's Federal Pension Insurance system, and the Benefits Data Trust in the United States [4].

Translating this book. Chapter 5-0 writes that this book was translated and proofread by communities, helped by translation tools across languages and cultures of the kind discussed in chapter 5-5 [5]. Hirokazu Nishio, who led the community behind the Japanese translation, has recorded that Takahiro Anno came across the idea of "broad listening" by reading the machine translation posted in the Japanese translation community (in Japanese) [6]. → Plurality in Japan

In more detail

What administration is

The book describes administration and bureaucracy as "structured forms of communication and rule-bound processing of this information that is much more formal and stricter than the conventions of natural language". Their aim is legitimacy, equity and procedural fairness. They govern much of law, property systems, identification, hiring and admissions: the core systems of states and companies [1].

ID cards, CVs, land deeds, tax returns, medical intake forms: according to the book, "Many of the most consequential junctures of life" turn on forms like these, which are "much thinner than the way we conduct most of our lives" [7].

Two opposite complaints

The book writes that complaints about bureaucracy come in two opposite kinds [8]:

1. Too rigid. Speed limits "are, most of the time, either too high or too low for the circumstances." People all over the world squeeze their lives into the format of CVs. In the late 1990s, a Dutch airline even ended up shredding hundreds of live squirrels that lacked the right paperwork at Schiphol airport [8].

2. Too complex. Like Franz Kafka's novel The Castle, bureaucracies are inscrutable, full of red tape and arbitrary discretion [8].

The book sums up that a key reason for the complexity is the need to handle a diversity of cases. The more a bureaucracy tries to span broad diversity, the more complex it becomes, until it loses legitimacy [8].

Old AI and today's AI

The "good old-fashioned AI" of the 1970s and 1980s tried to automate bureaucratic processing by encoding rules drawn from experts into nested sets: "Does the patient have a fever? If yes, are her eyes red...?" This ran into major obstacles [9].

Today's AI (especially GFMs) learns from many examples in a probabilistic way, and can reproduce the kind of flexible classification, response and reasoning that people show. The book sees this as opening the possibility of accepting inputs in many forms, without forcing specialist formats on users, while keeping a degree of consistency [9].

But the book also writes that "the power of GFMs stems less from being 'AI' than from their networked and probabilistic structure". The same kind of structure can exist in networks of human relationships (more adaptive forms of bureaucracy), it says [10].

Efforts already under way

Besides guidance on benefits, the book names efforts like these [4][10]:

services that help with legal advice
efforts to change hiring that favors graduates of elite schools; though the limited past examples of hiring diverse candidates can make such systems unreliable
work to map the traditional land rights of indigenous peoples and assert them in legal systems
beyond AI, collaborative mapping by Ushahidi, and distributed records used for refugee identities and land registries in Honduras

Frontiers

The book sketches two further prospects.

Next-generation badges. A proposal by Danielle Allen and others: gradually replace traditional coursework and grades with a range of "badges", from small ones recognizing specific skills up to larger ones, built from combinations of them, that employers can recognize. The book writes that AI could help employers read complex backgrounds, and students navigate diverse ways of learning [11].

Translating customs. Just as AI can cheaply translate languages, if it could also translate cultural rules, it might connect the diverse traditions overwritten by modernization and colonization (such as ways of settling family disputes, obligations of gift-giving, and uses of land) with today's legal systems, the book writes [12].

What is debated

It could make things harder to see. The book itself names the dangers: the opacity of GFMs; the "mystique of autonomy" implied by the word "AI" (which is why the book mostly avoids it); inheriting the biases of source data and creators; and potential misuse. Used in administration, the book writes, it might make dealings easier while deepening the opacity of bureaucracy, and do little to reduce arbitrary discretion or bias [13].
On whose terms is diversity taken in? The power relations under which diversity is absorbed into AI decide whether it opens opportunity or forces conformity. As a warning, the book notes that "Many of the ethnographers of old became tools of colonial subjugation rather than voices of inclusive translation" [13].
Abuse of legal gaps. The book cautions that "if abused by powerful interests, interoperability across legal regimes can easily slip into regulatory arbitrage, taking advantage of the gap between legal intent and formal rules" [13].

Where it fits in Plurality

Part 5 of the book sets out seven forms of collaboration, and Adaptive Administration is the fifth. It is more structured and reaches more widely than discussion (Augmented Deliberation), yet it allows longer communication than voting (⿻ Voting) [1][2].

It can also be seen as the counter at which people actually use the identity, property and access systems of Part 4. Chapter 4-1, too, signals that it will return in chapter 5-5 to GFMs adapted to users' relationships and contexts, as one way to handle the complexity of plural identity [14]. → Identity and Personhood, digital public infrastructure

The book also writes that other chapters make up for its weaknesses. Immersive Shared Reality and Post-Symbolic Communication could build trust between people that lets them hand over decisions, and methods of deliberation and voting could set the right distribution of power over how AI is run and steered [13]. → Civic AI

Sources

Plurality, 5-5, the two paragraphs after the break following the opening — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, chapter summaries (Chapter 5-5) — https://github.com/pluralitybook/plurality/blob/86158859464aee75633acd854c656928121a7fd8/contents/english/8-blurb.md ［CC0］
Plurality, 5-5, first three paragraphs — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 5-5, section "Adaptive administration tomorrow", bulleted list — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 5-0, opening — https://www.plurality.net/v/chapters/5-0/eng/ ［CC0］
Hirokazu Nishio, "Plurality in Japan（サイボウズラボ）" (Cosense, in Japanese) — https://scrapbox.io/nishio/Plurality_in_Japan%28%E3%82%B5%E3%82%A4%E3%83%9C%E3%82%A6%E3%82%BA%E3%83%A9%E3%83%9C%29
Plurality, 5-5, start of the section "Administration today" — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Same, the paragraphs on rigidity and complexity — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 5-5, section "Adaptive administration tomorrow", paragraphs 1–3 — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Same, last paragraph — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 5-5, section "Frontiers of adaptive administration" — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Same, the paragraph on translating customs — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 5-5, section "Limits of adaptive administration" — https://www.plurality.net/v/chapters/5-5/eng/ ［CC0］
Plurality, 4-1 "Identity and Personhood" — https://www.plurality.net/v/chapters/4-1/eng/ ［CC0］

We welcome error reports and questions ("What is…?"). Submissions become public issues.

Report an error in this article ・ Ask a question