Tech & society
The algorithm and the marabout
In Senegal, new technologies don't erase older ways of seeking advice.
Ndiaga Ndiaye4 min read
Tech & societyShare on LinkedIn
Tech & society
In Senegal, new technologies don't erase older ways of seeking advice.
Someone is facing an important decision.
They talk to their family.
Ask someone they respect.
Perhaps a religious guide.
A few hours later, they open an AI app.
Rephrase the same question.
And wait for a reply.
The scene is imaginary, but it says something about our time.
We haven't necessarily replaced older ways of asking for advice.
We've added a chat window to them.
In Senegal, the word marabout covers different realities, including religious teachers and spiritual guides. It doesn't describe one single practice, and certainly isn't just a synonym for predicting the future.
An algorithm, meanwhile, is a set of instructions. An AI system may generate answers through models trained on large collections of data.
These aren't the same kinds of knowledge or equivalent forms of authority.
Still, they raise a question that interests me: why do we trust a particular answer?
Because we know the person giving it?
Because other people respect them?
Because the wording sounds confident?
Or simply because the answer arrived quickly?
Ask an AI assistant a question and you may get a beautifully presented response.
Paragraphs.
Explanations.
Recommendations.
Delivered with remarkable confidence, even when important information is missing.
That's what fascinates me, and worries me a little.
A system that can produce a convincing sentence hasn't necessarily verified the claim behind it.
And a person receiving a clear answer may easily confuse clarity with reliability.
I use these tools in my work. I know how useful they can be.
But I try to remember one rule: a fluent answer isn't evidence.
In families and communities, trust is often built through relationships, experience and reputation.
You know who you're speaking to.
You can ask another question.
You may consider that person's history, responsibilities and role.
A digital system is different.
We don't always know which data was used, which design choices shaped the model or where its limits lie.
We see the output.
Not everything that produced it.
That doesn't make human advice automatically correct or digital advice automatically wrong.
It means the reasons for trusting either should remain open to examination.
It would be easy to contrast a supposedly "traditional" Africa with a "modern" technological world.
I find that contrast lazy.
The same person can use mobile banking, attend a religious gathering, browse social media and ask an AI to rewrite a letter.
There's no contradiction.
That's everyday life.
Technology doesn't arrive on a blank page. It enters existing habits, languages, relationships and ways of making decisions.
That's one reason language technology interests me. I wrote about it in Do you speak Wolof?.
An assistant that can't understand how someone actually speaks may be far less useful than it appears in a polished demo.
For practical tasks, I find it wonderful to ask a machine to compare options, explain a concept or help organise an idea.
But some answers concern health, money, law, family or deeply personal decisions.
In those cases, speed shouldn't be the main criterion.
Sometimes we need verifiable sources.
Sometimes a qualified professional.
Sometimes a real conversation with someone who understands our circumstances.
And sometimes simply the freedom to take time and think.
Technology can help with that process.
It shouldn't make decisions for us by pretending to be an authority it isn't.
I don't think the future will be a battle between machines and traditions.
I think it will be full of combinations, negotiations and new habits.
We'll keep asking for advice.
We may change who we ask.
And we'll need to learn the difference between an answer that helps us think and one that merely reassures us.
The question isn't only who gives the answer.
It's what gives us a good reason to believe it.
Even when it appears inside a very pretty chat bubble.
