Annette Baier and slow philosophy: a proposal the age of AI has revived

Some proposals are unintelligible at the moment they are made and unavoidable thirty years later.

Annette Baier's proposal about academic assessment is one of them.

Born in New Zealand, Baier (1929-2012) taught for twenty-four years at the University of Pittsburgh and retired in 1997. She was among the most important Hume scholars of her generation; her most durable contribution, though, lies elsewhere.

Trust: the concept moral philosophy forgot#

Baier's 1986 paper "Trust and Antitrust" opened a field in contemporary moral philosophy.

The observation behind it was simple and uncomfortable. For centuries moral philosophy had worked from a model of contract between equal and autonomous adults. Duty in Kant, calculation in utilitarianism, agreement in contractarianism — every one of them assumed parties independent of one another.

Yet most of human life does not fit that model.

There is no contract between an infant and its parent. There is no equality between patient and doctor, student and teacher, an old person and the person who cares for them. What governs these relationships is trust — and trust is by definition an unequal relationship: the party who trusts makes herself defenceless.

Baier's question was why moral philosophy had set such relationships aside and examined only agreement between equals.

Her answer was equally plain. Because the people who wrote the philosophy were men who had handed the part of life that requires care to somebody else — most often to women.

The anatomy of trust#

Baier did not use trust merely as a critical cudgel. She analysed it.

Trust, on her account, is a three-place relation: A trusts B with respect to X. Trust always concerns some particular domain. You may entrust your child to someone and not entrust them with your money.

Her second distinction cuts deeper. Trust and trustworthiness are not the same thing. That a relationship contains trust does not show that the relationship is morally good. Baier's criterion is famous:

A relationship of trust is moral if it can survive the disclosure to each party of how the other is trusting.

Relationships built on exploitation fail this test. The exploiting party loses the other's trust the moment the real structure of the relationship becomes visible.

The criterion is in direct use today in arguments about institutional trust, digital platforms and algorithmic systems: can a system sustain trust once it explains to the user how it works?

Reading Hume#

Baier's work on Hume belongs to the same line.

A Progress of Sentiments (1991) reads Hume's A Treatise of Human Nature as a whole — which is not the usual practice; Hume is generally remembered for the epistemology.

Baier's thesis is that Hume's book, having shown that reason cannot ground itself, goes on to say what one does next. The answer is sentiment, habit and the practice of living together. Hume's scepticism, in other words, is not a dead end but a door opening on to a social ethics.

Last week this column marked the anniversary of Hume's death. Baier's reading is the modern continuation of that line.

The "slow philosophy" proposal#

Baier's proposal to the academy grew out of a single sentence in Moral Prejudices, set in parentheses and followed by a question mark.

It runs like this: in tenure and promotion decisions, assess the candidate not on the whole list of publications but on four pieces of their own choosing.

In 2015 Jennifer Whiting of Pittsburgh — a colleague of Baier's for a decade — argued the case for it at length in Daily Nous. The piece was titled "A Modest Proposal: Slow Philosophy".

The reasoning was this. A criterion of quantity rewards precisely what is wrong in philosophy. If a philosopher publishes four articles a year, the chances are that none of them has been thought about for long enough. Most of the transformative texts in the history of the discipline are single essays worked at for years.

The proposal drew interest at the time. Nothing came of it.

Why it has come back now#

On 19 August, Daily Nous put the proposal back on the table. The reason is technological.

Generative artificial intelligence has cut the cost of producing academic text to a fraction of what it was. Preparing a draft, assembling a bibliography, writing a summary of the literature: these take minutes now, not hours.

The consequence follows at once: the number of publications is ceasing to be an index of labour.

The quantitative criterion was troublesome enough already; it has now lost its connection to whatever it claimed to measure. If a candidate has forty publications, that shows neither industry nor depth.

The journal Philosophy & Public Affairs banning artificial intelligence authorship, which we reported this month, and the row over accusations that followed it, were the publishing side of the same problem. Baier's proposal looks at the assessment side.

And her remedy does not depend on detection tools. It simply changes the criterion: ask not how many pieces a person has written but what their best four are.

Artificial intelligence does not make it easier to produce four good papers. It makes it easier to produce forty mediocre ones.

What the proposal owes to Baier's own life#

That the proposal came from Baier is no accident.

She began her academic career late and wrote little, at length considered. Her body of work consists largely of essays; even her books are collections of essays. "Trust and Antitrust" is a single paper, and it opened a field.

The proposal was also part of a wider criticism of the career timetable imposed on women academics — an aspect that was misreported at the time, with a proposal that was not hers attributed to her.

Whiting's 2015 piece set out, among other things, to correct that misreporting.

What it says today#

Baier's two contributions rest on a single intuition.

On trust: moral life is not exhausted by agreement between equals; defencelessness and dependence are subjects of morality too.

On slow philosophy: intellectual worth cannot be reduced to countable output; some things take time.

Both refuse the same thing: the assumption that human relationships and human thought can be converted into calculable units.

It is not hard to see why that refusal is being read again in a period when measurement and automation are reaching into every domain.

In Turkish#

Baier's books have not been translated into Turkish. "Trust and Antitrust" is among the texts most frequently cited in the Turkish-language literature on the ethics of trust.