Say you need a spreadsheet total you can trust. You could watch a lesson on a new formula feature. Or you could learn how to organise the data. Which would help more?
If you understand the data but can’t enter the formula, the formula may be what you need next. If you don’t know whether some rows are repeated, finding the right command won’t answer that question.
The choice depends on what is stopping you from doing the work. It also depends on where you could use what you learn again. A lesson about a general idea won’t always help in many places. A specific procedure can be worth learning when it solves the problem in front of you.
This module is for working adults choosing what to learn for tasks they do again and again, where the same demands come up across tools or situations.
What you’ll get from this module
In this module
By the end of this module, you’ll have a clearer way to choose what to learn next. We’ll look at research on how learning carries over to new tasks, on learning to program, on software training, on AI help at work and on money training for small businesses. You’ll connect those findings to your own work, pick a situation from daily life and write down a learning decision. You’ll finish with a next step and a reason for it, or a clear question to answer first.
Ask Jing
Become a Jingius.
What makes learning useful beyond the lesson?
A research review: say where the learning needs to go.
Susan M. Barnett and Stephen J. Ceci’s 2002 review looked at transfer: using what you learned in one situation in another. Their framework keeps two things apart. One is what you learn. The other is where and when you use it.
Using a skill on a different subject, in a new place, at a later time or with other people can ask different things of you. A task can be like the first one in one way and unlike it in another. So “useful everywhere” is a much bigger claim than “useful on a similar task.”
The review helps describe those differences. It doesn’t give one overall chance that learning will carry over. Instead, the useful question gets more concrete: which part of what you learned could help with which other task?
Why it matters: A broad label doesn’t show where a skill will help. Calling a lesson “reasoning” tells you less than saying what you need to think through in two particular situations.
Read the review on when and where learning transfers.
A review of many studies: learning can carry over, but the comparison matters.
Ronny Scherer, Fazilat Siddiq and Bárbara Sánchez Viveros looked at teaching programming in their 2019 meta-analysis. A meta-analysis pools the results of many studies. This one covered 105 studies and 539 effect sizes. An effect size measures how far apart the groups being compared ended up. The studies looked at learning to program and at other thinking skills.
The review found gains on programming tasks and on tests of other thinking skills. But the gains were smaller when the comparison group had a different kind of training aimed at the skill being tested. Other comparison groups had no such targeted training, though they might still have had their usual lessons. Studies that were never published also showed smaller gains than published ones.
These findings show why we shouldn’t rule out learning carrying over. The results were positive. But the size of the gain depended on what was measured and what the other group did.
Why it matters: That gives you a reason to look for other tasks where learning might help. It doesn’t make programming the right next lesson for every worker. You still need to link the lesson to work you actually do.
Read the meta-analysis of programming and transfer.
A workplace experiment: teaching the ideas didn’t give the expected edge.
Lorne Olfman and Munir Mandviwalla’s 1994 study compared two ways of teaching university staff to use Windows 3.0. One focused on ideas: how the screen and its parts worked. The other focused on steps: how to get tasks done.
Over four sessions, the researchers tested what people learned. Later they asked how people were using the software. The ideas-first approach did not show the edge the researchers expected over the steps-first approach. Fewer people turned up as the study went on, and access to the software was limited.
This result challenges the idea that learning the ideas is always the better use of your time. It doesn’t show that the two approaches work equally well everywhere.
Why it matters: You may need both: an understanding of why something works and a set of steps for doing it. A procedure isn’t automatically a distraction from learning something worthwhile.
Read the field experiment on conceptual and procedural software training.
A workplace study: a specific tool can help with the work in front of you.
Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied an AI assistant that one customer-support firm gave to its workers at different times. Their November 2024 paper uses data from 5,172 support agents.
Having the assistant went with an estimated 15% more issues solved per hour, on average. Less experienced or lower-performing agents gained more. The researchers watched how the tool was rolled out at work. Workers had a short introductory training before they got access.
The estimate describes how people worked while the assistant was there, in that workplace. It doesn’t single out the effect of a training session. Nor does it show that workers kept such a large gain after losing the tool.
Why it matters: That leaves a useful finding in place: a particular tool can help with particular work. Learning to use it may come first, even if you expect to use it little anywhere else.
Read the manuscript on generative AI at work.
A randomised training trial: useful basics can take a simple, practical form.
Alejandro Drexler, Greg Fischer and Antoinette Schoar compared two training offers for very small business owners who were clients of ADOPEM in the Dominican Republic. Their paper reports a trial with 1,193 clients who had said they wanted training. Clients were picked at random for a training offer or for a group with no training.
One programme taught standard accounting. The other taught simpler money rules. One habit in the simpler programme was to keep business money and household money apart, and to write it down whenever money moved between the two.
Compared with the group offered no training, the simpler programme raised the overall score for good money habits and cut mistakes in reporting. The standard accounting training made no difference to these two overall measures that the researchers could be confident of.
Among owners with fewer skills to start with, the simpler programme cut reporting mistakes more than the accounting training did. That gives a concrete reason to think about which approach suits the people taking the course.
The researchers tested the whole training programme. They couldn’t tell how much of its effect came from the money-separation habit alone. The result does show why the way something is taught matters when you choose what to learn.
Why it matters: The money-separation habit made the line between business and household money concrete. A useful idea underneath can be taught through a procedure people can follow. Ask what will help you do the task, rather than assuming a more abstract lesson is better.
Read the manuscript comparing financial training approaches.
What this means for you
Choose what to learn by the work it can help you do. Think about where else you could use it, but leave room for a procedure that solves an important problem right now.
Separate the judgment from the operation
Describe the result you need without naming a product. “Produce a total I can trust” says what the work must do. “Learn the new spreadsheet feature” names one thing to learn before you know it’s needed.
Then split what you need to understand or decide from what you need to do in the software. In the made-up total, knowing what each row stands for and checking for repeated rows are judgments about the data. Entering the formula is the action you take in the spreadsheet.
A lesson can hold both. Learning the formula may also teach you which values belong in the total. Keep that useful link. Don’t file the whole lesson as either shallow or fundamental.
Ask what the operation does and when it’s the right one to use. If that’s clear and the command is what’s in the way, learn the command. If the judgment is unclear, learn the reasoning you need to use the command well.
Name the next use before calling a skill transferable
Name another task, tool or situation where the learning could help. Say what the two tasks would ask you to do in the same way. A second product name on its own doesn’t show that the task underneath is alike.
In the spreadsheet example, you might look at the same data in another program. The question is whether your thinking about rows and repeats still holds, even if you have to write the formula a different way.
When you can, try a similar task in that other place. If you’ve only pictured the second use, treat it as something you expect, not something you know. One success tells you about that situation. It doesn’t prove the skill is useful everywhere.
These working labels help describe each part of what you’re learning. They aren’t rankings of whole tools or people:
| Working label | What you mean | Boundary to record |
|---|---|---|
| Transferable fundamental | A skill you have used, or expect to use, in more than one named situation | Name the situations and distinguish what you have tried from what you expect. |
| Domain fundamental | A rule or judgment underlying recurring tasks in one area of work | Name the area and the kinds of tasks it serves. |
| Tool feature | An action tied to a particular product, interface or setup | State the current use; it can still have considerable immediate value. |
The labels can overlap. A rule that several tools share may help across all of them. A skill from one area of work may need changing when the task changes. Update your description when those things change.
For example, checking for repeated rows is part of working with tables of data. If you’ve used that judgment on similar data somewhere else, you can also say where it carried over. Both descriptions can fit. Name the other situation, so it’s clear what you’ve actually tried.
Put the current obstacle in its proper place
If you can’t deliver an important task right now, spend your next session on the least you need to learn to get past that obstacle. Include a feature if the task needs it.
Once you can deliver the work, favour what you need again and again in your field, and skills you’ve already used in other relevant places. Put these ahead of optional features with no named use.
A feature with a clear payoff in your own work can still be worth learning. Weigh what it’s likely to give you while you use it against the effort of learning it and keeping up with it.
If the future payoff rests only on a claim that a skill is useful everywhere, hold off deciding. Name a likely place you’d use it before you commit a large block of learning time.
Go back to the made-up spreadsheet choice. If you don’t know how to write the formula, learn that. If you don’t know what the rows stand for, work on understanding the data. The next lesson changes because the obstacle changes, even though the result you want stays the same.
Keep the decision tied to a real requirement
Say you want to write a request people can follow. You need to decide who will read it, what decision you need from them and what information will help them make it. Finding a formatting command is a separate part of the work.
If the look of the page is what’s really needed, a lesson on formatting may be the useful next step. If the reader can’t tell what decision you need, extra formatting won’t fix that.
Writing a different kind of document, or in another language, can change what you need to know. When you use search or AI, decide on your question and what a good enough answer must show. Don’t assume one prompt works across different models or tasks.
Look at your next learning choice again when the work changes. Let real need and the benefit you’ve seen decide what gets more attention, not how new a release is.
How about your daily life?
Where are you deciding between learning how to do something in a tool and understanding the work behind it?
You could pick a spreadsheet task you do often, a request you often write, or a question you answer with help from search. Look for something where you can describe the result you want and a possible second use.
Write down situations you’d like to explore. One good idea is enough.
| A situation in my daily life | What I want to accomplish or understand |
|---|---|
| ____________________ | ____________________ |
| ____________________ | ____________________ |
Which situation would be most useful to work through now?
My chosen situation: ____________________
Take this situation into the exercise below. You can change your mind. Picking it doesn’t yet decide what you should learn first.
Choose your next learning step
Start with your chosen situation. If you came straight here, pick a work task you do again and again. You should be able to say what it needs and where those needs might come up again, with another tool or in another situation.
- Name the output and the obstacle. Describe what you need to get done without naming a product. Say what is stopping you from delivering it right now.
Output and current obstacle: ____________________ - Separate the learning components. Write down the rule, judgment or skill involved. Then write the action in the software that carries it out. Say what that action does and when it’s the right one to use. A lesson may hold both parts.
Judgment or skill: ____________________
Operation and when it applies: ____________________ - Name a credible next use. Name another task, tool or situation, and say what would need to stay the same. If you can, try a similar task there. Write down what you saw, or mark the second use as something you expect.
Destination, shared demands and observed or expected use: ____________________
If you can’t name another place to use it, leave wider use as an open question. One success doesn’t show that a skill carries over widely. - Describe each part and where it helps. Use “transferable fundamental” for a skill you’ve used, or expect to use, in more than one named situation. Use “domain fundamental” for rules or judgments you need again and again in a named area of work. Use “tool feature” for an action tied to a particular product or screen. The labels can overlap. Update the description when things change.
Component, description and boundary: ____________________ - Choose the next session. If an important task is stuck right now, learn the least you need to get past the obstacle, including a feature if you must.
Once you can deliver it, favour skills you need again and again in your field, or have already used elsewhere, over optional features with no named use. Learn a useful feature when what it’s likely to give you is worth the effort of learning it and keeping up with it.
If all you have is a vague promise that something is useful everywhere, hold off deciding. Name the task or information that’s missing before you commit a large block of time.
Next learning step, reason or question to resolve: ____________________ - State when to reconsider. Look at the choice again when the task, or how you expect to use the learning, changes. Compare what you really needed and gained with what you expected.
Change that would prompt a review: ____________________
Ask Jing
Become a Jingius.
Your next step
You’re finished when you have one next learning step with a reason based on your task, or a clear gap to fill first. Keep a note of where else the learning could help and what would need to stay the same.
A note on the evidence
The studies look at different kinds of learning and work. They cover learning carrying over to new tasks, teaching software, working with AI help and training owners of very small businesses. These questions are related, but they aren’t the same.
The Windows result challenges the claim that teaching ideas is always better. The AI estimate comes from one firm’s rollout, and the money-training result is about a whole training offer. Those differences matter when you judge how closely a result fits your situation.
How we prepared this module
This module draws on a review of transfer, a meta-analysis on programming, and studies of software training, help at work and money training. Made-up spreadsheet and writing examples explain the decisions. The written exercise helps you apply those ideas to one task you do again and again.
Sources
- Susan M. Barnett and Stephen J. Ceci. (2002). When and Where Do We Apply What We Learn? A Taxonomy for Far Transfer. Psychological Bulletin, 128(4), 612–637. Published article; pp. 621–623, content and context taxonomy.
- Ronny Scherer, Fazilat Siddiq and Bárbara Sánchez Viveros. (2019; online 25 October 2018). The Cognitive Benefits of Learning Computer Programming: A Meta-Analysis of Transfer Effects. Journal of Educational Psychology, 111(5), 764–792. Published article; p. 764 abstract; pp. 769–770, methods and control groups; pp. 777 and 784, transfer and comparison results.
- Lorne Olfman and Munir Mandviwalla. (December 1994). Conceptual Versus Procedural Software Training for Graphical User Interfaces: A Longitudinal Field Experiment. MIS Quarterly, 18(4), 405–426. Author-uploaded published text; pp. 409–410 Research Design and Subjects, p. 415 results and p. 420 conclusion.
- Erik Brynjolfsson, Danielle Li and Lindsey Raymond. (May 2025; online 4 February 2025). Generative AI at Work. Quarterly Journal of Economics, 140(2), 889–942. November 7, 2024 arXiv v2 manuscript used here; abstract and section 3.1, “AI Rollout.”
- Alejandro Drexler, Greg Fischer and Antoinette Schoar. (April 2014). Keeping It Simple: Financial Literacy and Rules of Thumb. American Economic Journal: Applied Economics, 6(2), 1–31. Author manuscript
KIS-DFS-March2013.pdfused here; pp. 2–4, training and sample; pp. 13–17 and Tables 2, 4 and 6, outcome and subgroup comparisons.
