One way to survey what computing does is to follow it through widening circles: a classroom, the school that runs the classroom, the homes students go back to, the community around them, and finally the broader society. The tools change from circle to circle, but a pattern recurs. Each does something that used to be difficult, and each has a limit that its appearance can hide, so the person using it has to understand what it actually does before relying on it.

In the classroom#

Many classes now run partly on a learning management system (LMS), a platform that distributes materials, collects work, and records activity. Because those records are kept student by student, its reports can show a teacher who needs support: missing work and quiz results are evidence of what a particular student has or hasn't done. Activity counts such as logins are a different kind of data. They measure use of the platform, not learning, so a student with few logins and strong work is no contradiction, and an administrator can't judge the quality of instruction from login totals. Opening a video shows access and nothing more about understanding. Treat these counts as a prompt to investigate, perhaps by asking the student what they took from the videos, rather than as evidence of understanding.

An LMS keeps track of what students do, while two other kinds of tools change what students encounter. Adaptive software changes content based on a learner's responses, choosing the next problem according to how the last one went. Assistive technology widens access to content: captions present a recorded lecture as text, and text-to-speech lets a student who comprehends better by ear hear a reading. So adaptive tools respond to performance, and assistive tools remove barriers. Both still depend on a teacher's judgment about whether they serve the learning goal, because software doesn't know your goals. An adaptive program can move a student smoothly through a sequence that has little to do with your course, which is why you review its data and ask whether the content it selects fits what you are teaching.

In the school#

Classroom tools sit on top of systems that the school runs as an institution, and one of these, the network, shapes what a teacher can do on a given day. On a school-managed network, the IT department controls access. Its firewall (see 0003.d) applies rules that allow or block outside addresses as traffic crosses the network's boundary, and the staff who manage it maintain those lists. A teacher whose class can't reach a resource therefore asks IT or the network administrator, ahead of the lesson, rather than bypassing the filter. The diagnosis is often simple: if a site works at home on the same laptop but not at school, only the network has changed, so the school's filtering is the likely cause. The request needn't be all or nothing, since allowing specific approved sources can meet a teaching need while keeping a broader restriction in place.

Schools also gather a great deal of data, and data is only useful if it's right. Digital forms promote accuracy through validation, which means checking data types and ranges at entry, so that errors can be fixed immediately or set aside (isolated) for review. A field that accepts only whole numbers from 2 through 5 enforces a type and a range at once. Drop-down lists go further by restricting values to a fixed set, which keeps data consistent, so nobody has to reconcile "9th," "nine," and "Grade 9" later. Validation has limits, though, and they show what it is really doing. A format check can't detect a well-formed address that doesn't exist, while checking against reference data, such as a list of real addresses, can. Other improvements that are sometimes offered for this problem address something else: encryption, access controls, and faster databases protect or speed data, but they don't make it accurate.

Schools also need to talk with families, including multilingual families, and machine translation speeds that communication considerably. Its weakness is one that's hard to see: a translation can miss context and meaning while looking fluent, so a reader can't judge its quality from its polish. The sensible practice is to match review to the stakes, which depend on the consequences of an error. A quick reminder may need little review, and a bilingual volunteer can check routine notices, while emergency procedures or special education meetings need a fluent person or a qualified interpreter.

At home#

Those messages arrive in homes that have computing of their own. Internet of Things (IoT) devices, such as smart thermostats and speakers, add sensing, remote control, and automation to everyday objects. A washing machine that notifies its owner when a cycle has finished is doing something only a network connection makes possible. These devices usually send data to a company's servers, including spoken requests and usage records, and this dependence has two consequences. That data can reveal routines, such as when a home is empty, so fine-grained energy use says more than the size of a bill. And many remote features stop working if the company shuts down its servers, even though the hardware itself remains intact.

All of this assumes a connection, which some homes don't have. Satellite service reaches remote areas where laying cable is impractical, and that reach is its main advantage. Where wired service exists, wired usually offers more speed and less latency (delay) for the price, so satellite makes the most sense when students need connectivity now and no wired option is available yet. The delay comes from distance. Geostationary satellites orbit far from Earth, and the long signal path up and back adds noticeable delay, so downloads work well, but interactive uses such as games and calls feel sluggish. Weather can also degrade the signal. Section 0003.g returns to latency and to low-Earth-orbit satellites.

In the community#

Beyond the home, computing gives groups of residents ways to build evidence that used to require an institution. Public data portals publish government data that anyone can analyze and reuse, low-cost sensors can fill gaps in official monitoring, and mapping tools make the results easy to share. Together they let communities gather and share evidence on air quality, transit, or housing for civic advocacy, for example by placing sensors near a busy road and mapping the readings, or by using a transit agency's published schedule data to compare how often buses serve different neighborhoods.¹ The evidence has to withstand challenge, though. Low-cost sensors can drift or read differently from official instruments, so calibrating them against a reference instrument makes the evidence more credible.

In the broader society#

Two technologies from the widest circle show how much depends on knowing what a tool is suited to. A 3-D printer builds an object layer by layer from a digital model, a process called additive manufacturing. Because no mold is required, it suits custom parts, prototypes that can be revised cheaply, and small batches, such as a replacement for a discontinued part. Mass production favors injection molding, which has high setup cost but low cost per unit at scale, since the expense of the mold is spread over many copies, and printing is slower per object. Being able to print a model doesn't settle whether you may sell the result: printed models are still subject to their licenses, a topic 0001.f takes up.

The second technology is the large language model (LLM), which generates text by predicting likely next words from patterns in its training data. It doesn't retrieve verified facts, and much of how it behaves follows from that. It is most reliable on topics well represented in that data, such as a concept from introductory biology, and weakest on recent, niche, or private information. It can state errors confidently and invent details, including citations, because a plausible-looking reference fits the pattern whether or not the source exists. Confident wording therefore tells you nothing about accuracy, and asking the model whether it is right only produces more generated text, so check claims against independent sources.² Both technologies, like the others in this section, have effects that go beyond the person using them, and those wider effects are the subject of 0001.b.

• learning management system (LMS): a platform that distributes materials, collects work, and records activity
• adaptive software: software that changes content based on a learner's responses
• assistive technology: tools that widen access to content, such as captions and text-to-speech
• Internet of Things (IoT): everyday objects with added sensing, remote control, and automation over a network
• firewall: a system whose rules allow or block traffic crossing a network boundary
• validation: checking data types and ranges as data is entered
• public data portal: a site that publishes government data for anyone to analyze and reuse
• geostationary satellite: a satellite in a distant orbit, which adds noticeable latency
• additive manufacturing: building an object layer by layer from a digital model
• large language model (LLM): a model that generates text by predicting likely next words from patterns in training data

Practice: practice questions for 0001.a

1. MWC1, Argument project: collecting and using data to make a claim. ↩

2. MWC2, Estimation: fitting models and overfitting. ↩