What is Machine Learning—and what should educators know about it?
In this AI Foundations video from Ed3, we explain machine learning in plain language, using a simple story (chickens + egg patterns) to show what ML really is: learning from data to make predictions.
Machine learning powers many of the tools educators see every day—adaptive learning platforms, personalization systems, plagiarism detectors, and the algorithms underneath generative AI. But it also has limits: ML learns patterns, not truth, and it can only be as fair and accurate as the data it’s trained on.
This video covers:
A common misconception is that machine learning is objective. It isn’t. It reflects patterns in the training data, including blind spots, biases, and omissions. That’s why understanding ML helps educators use tools critically—without outsourcing professional judgment.
This video is part of the AI Foundations series by Ed3, supporting educators worldwide in making informed, ethical, and human-centered decisions about AI in classrooms.
👉 Learn more about Ed3: https://www.ed3global.org
👉 Explore professional learning, courses, and events designed for educators navigating AI responsibly.
👉 Join our community of practice: https://community.ed3global.org
00:11 A simple story: learning patterns (chickens and eggs)
00:39 What “machine learning” means
00:47 How ML powers generative AI
00:54 Learning from examples vs coding rules
01:11 Where ML shows up in education
01:21 The blind spots: bias + missing data
01:50 What this means for students and teachers
02:00 Three ways teachers can use ML wisely
02:44 Machine learning = statistics at scale
02:53 Separating hype from reality
1
00:00:11,366 --> 00:00:13,366
A year ago, I bought some chickens
2
00:00:13,366 --> 00:00:15,366
and at first I had no idea
3
00:00:15,366 --> 00:00:16,833
when I should be collecting their eggs.
4
00:00:16,833 --> 00:00:18,300
So, I decided to check
5
00:00:18,300 --> 00:00:19,966
several times a day.
6
00:00:19,966 --> 00:00:20,966
Then I started to notice
7
00:00:20,966 --> 00:00:22,333
that the chickens would lay eggs
8
00:00:22,333 --> 00:00:24,333
at different times of the day.
9
00:00:24,333 --> 00:00:26,400
So I adjusted my egg collecting schedule
10
00:00:26,400 --> 00:00:28,433
to match what the chickens were doing.
11
00:00:28,433 --> 00:00:29,433
So essentially
12
00:00:29,433 --> 00:00:30,833
I was learning from the patterns
13
00:00:30,833 --> 00:00:32,200
I was observing.
14
00:00:33,400 --> 00:00:35,566
As humans, we're constantly learning
15
00:00:35,566 --> 00:00:37,600
based on what we experience
16
00:00:37,600 --> 00:00:39,500
and when machines try to be like humans
17
00:00:39,500 --> 00:00:40,733
and try to learn
18
00:00:40,733 --> 00:00:42,633
based on the experiences and data
19
00:00:42,633 --> 00:00:43,766
they’re collecting,
20
00:00:43,766 --> 00:00:45,866
we call that “machine learning”.
21
00:00:45,866 --> 00:00:47,133
Machine learning powers
22
00:00:47,133 --> 00:00:48,933
generative artificial intelligence
23
00:00:48,933 --> 00:00:50,466
by producing the algorithms
24
00:00:50,466 --> 00:00:53,400
that predict the next pixel or text.
25
00:00:53,400 --> 00:00:54,533
For example,
26
00:00:54,533 --> 00:00:56,766
instead of coding every grammar rule,
27
00:00:56,766 --> 00:00:59,766
we feed an AI thousands of sentences.
28
00:00:59,766 --> 00:01:00,533
Over time,
29
00:01:00,533 --> 00:01:01,566
it learns to predict
30
00:01:01,566 --> 00:01:04,300
the next logical word in the sentence.
31
00:01:04,300 --> 00:01:05,000
The system isn't
32
00:01:05,000 --> 00:01:06,466
memorizing language,
33
00:01:06,466 --> 00:01:08,200
it's building statistical models
34
00:01:08,200 --> 00:01:09,400
to make increasingly
35
00:01:09,400 --> 00:01:10,500
accurate predictions.
36
00:01:11,600 --> 00:01:13,700
Machine learning is everywhere,
37
00:01:13,700 --> 00:01:15,833
including adaptive math platforms,
38
00:01:15,833 --> 00:01:17,333
plagiarism detectors,
39
00:01:17,333 --> 00:01:19,600
and personalized reading apps.
40
00:01:19,600 --> 00:01:21,733
But it also has blind spots.
41
00:01:21,733 --> 00:01:23,100
If a reading app is trained
42
00:01:23,100 --> 00:01:25,133
mostly on Western literature,
43
00:01:25,133 --> 00:01:26,466
it may struggle to recognize
44
00:01:26,466 --> 00:01:28,466
dialects, non-standard grammar,
45
00:01:28,466 --> 00:01:30,800
or culturally diverse examples.
46
00:01:30,800 --> 00:01:32,333
This means that the intelligence
47
00:01:32,333 --> 00:01:33,933
is only as good as the data
48
00:01:33,933 --> 00:01:35,266
it is trained on.
49
00:01:35,266 --> 00:01:36,666
That's why it's so important for us
50
00:01:36,666 --> 00:01:38,766
to understand how machine learning works,
51
00:01:38,766 --> 00:01:40,233
so we can use it critically
52
00:01:40,233 --> 00:01:41,600
and not blindly.
53
00:01:41,600 --> 00:01:42,766
For students,
54
00:01:42,766 --> 00:01:44,100
this could mean an adaptive app
55
00:01:44,100 --> 00:01:45,500
that helps most kids,
56
00:01:45,500 --> 00:01:47,333
but struggles when a learners style
57
00:01:47,333 --> 00:01:49,233
or identity isn't well represented
58
00:01:49,233 --> 00:01:50,833
in the training data.
59
00:01:50,833 --> 00:01:52,133
For teachers,
60
00:01:52,133 --> 00:01:53,266
it means we need to understand
61
00:01:53,266 --> 00:01:54,400
the limits.
62
00:01:54,400 --> 00:01:56,600
AI is learning from patterns,
63
00:01:56,600 --> 00:01:59,566
not from truth or lived experience.
64
00:01:59,566 --> 00:02:00,733
There are three things
65
00:02:00,733 --> 00:02:01,833
teachers can do to use
66
00:02:01,833 --> 00:02:03,433
machine learning wisely.
67
00:02:03,433 --> 00:02:04,500
Number one.
68
00:02:04,500 --> 00:02:06,733
Stay curious about the data.
69
00:02:06,733 --> 00:02:08,066
Ask yourself
70
00:02:08,066 --> 00:02:09,266
“what kind of data did this”
71
00:02:09,266 --> 00:02:10,166
“tool learn from”
72
00:02:10,166 --> 00:02:12,300
“and doesn't reflect my students?”
73
00:02:12,300 --> 00:02:15,000
Number two, use AI as a draft,
74
00:02:15,000 --> 00:02:16,533
not a decision.
75
00:02:16,533 --> 00:02:18,600
Machine learning can speed up feedback,
76
00:02:18,600 --> 00:02:19,833
but your professional judgment
77
00:02:19,833 --> 00:02:22,100
ensures it's accurate and fair.
78
00:02:22,100 --> 00:02:23,233
And number three,
79
00:02:23,233 --> 00:02:25,433
turn it into a learning moment.
80
00:02:25,433 --> 00:02:26,600
Use machine learning
81
00:02:26,600 --> 00:02:28,900
as a conversation starter with students
82
00:02:28,900 --> 00:02:31,200
about how knowledge is shaped by data,
83
00:02:31,200 --> 00:02:32,900
and whose data is missing.
84
00:02:32,900 --> 00:02:34,033
You could even show students
85
00:02:34,033 --> 00:02:36,133
how pattern recognition works.
86
00:02:36,133 --> 00:02:37,700
Let them train a simple model
87
00:02:37,700 --> 00:02:39,633
by sorting pictures or words
88
00:02:39,633 --> 00:02:41,166
so they see both the power
89
00:02:41,166 --> 00:02:42,933
and the pitfalls.
90
00:02:42,933 --> 00:02:44,933
See, machine learning isn't magic,
91
00:02:44,933 --> 00:02:47,133
it's statistics at scale.
92
00:02:47,133 --> 00:02:48,633
Pattern recognition that improves
93
00:02:48,633 --> 00:02:50,133
as it gets exposed to more,
94
00:02:50,133 --> 00:02:52,433
and hopefully better data.
95
00:02:52,433 --> 00:02:53,800
As educators,
96
00:02:53,800 --> 00:02:54,766
Knowing how it works
97
00:02:54,766 --> 00:02:55,466
helps us separate
98
00:02:55,466 --> 00:02:57,633
the hype from the reality
99
00:02:57,633 --> 00:02:58,566
so we can make wiser
100
00:02:58,566 --> 00:03:00,033
choices for our classrooms.
