The AI Education Divide: How Technology Access is Reshaping Australia’s Future Workforce

Five data visualisations exploring how unequal digital foundations in Australian schools are compounding into unequal outcomes as AI reshapes study, work, and pay.

Australia’s digital literacy is going backwards, not forwards

Every three years since 2005, the Australian Curriculum, Assessment and Reporting Authority has tested a national sample of Year 6 and Year 10 students on their ability to actually use digital technology: creating presentations, analysing data, judging whether a website can be trusted, recognising a phishing attempt. The chart below tracks the share of students who cleared the “Proficient Standard” - a bar the testing body itself describes as a “challenging but reasonable” expectation for that year level - across every cycle for which national results have been published. The line rose through the early 2010s as school laptop and tablet programs expanded, plateaued through the mid-2010s, and has now fallen in both year levels at once, arriving at its lowest point in the programme’s twenty-year history in the same testing cycle that followed the mainstream arrival of generative AI in classrooms.

# SOURCE: Australian Curriculum, Assessment and Reporting Authority (ACARA).
# NAP-ICT Literacy Public Reports, 2005-2025 cycles (nap.edu.au).
# Percentage of students at/above the "Proficient Standard".
# Figures below are the ONLY cycle results publicly confirmed in ACARA/ACER
# reporting; missing cycles are left as NA rather than estimated.
ictl <- tribble(
  ~year, ~year_level, ~pct_proficient,
  2005,  "Year 6",   49,
  2011,  "Year 6",   62,
  2017,  "Year 6",   53,
  2022,  "Year 6",   55,
  2025,  "Year 6",   50,
  2017,  "Year 10",  54,
  2022,  "Year 10",  46,
  2025,  "Year 10",  37
)
p1 <- ggplot(ictl, aes(x = year, y = pct_proficient, colour = year_level,
                        group = year_level,
                        text = paste0(year_level, ", ", year, ": ",
                                      pct_proficient, "% proficient"))) +
  geom_line(linewidth = 1) +
  geom_point(size = 2.5) +
  scale_colour_manual(values = c("Year 6" = cb_purple, "Year 10" = cb_orange)) +
  scale_y_continuous(limits = c(0, 70), labels = label_percent(scale = 1)) +
  scale_x_continuous(breaks = c(2005, 2011, 2017, 2022, 2025)) +
  labs(
    subtitle = "% at/above the NAP-ICT Literacy Proficient Standard",
    x = NULL, y = NULL,
    caption = "Source: ACARA, NAP-ICT Literacy Public Reports (2005-2025), nap.edu.au. CC BY 4.0."
  )

ggplotly(p1, tooltip = "text") %>%
  layout(legend = list(orientation = "h", y = 1.15), margin = list(t = 50))

The most striking feature of this chart isn’t just that scores fell - it’s when they fell furthest. The Year 10 line drops nine percentage points between 2022 and 2025 alone, a steeper single-cycle decline than at any earlier point in the twenty-year series, and 2025 marked the first time this specific cohort of students had been tested since generative AI tools became a fixture of daily school life. That timing doesn’t prove causation on its own, but it sits uncomfortably alongside a pattern researchers analysing the same results have already flagged: students are using digital tools constantly, but that constant use is not translating into the underlying literacy the tools were meant to build on top of. Everything that follows in this report is, in one way or another, a question about who gets left holding that gap.

Digital disadvantage doesn’t arrive as a single gap - it compounds across access, cost, and skill

If the previous chart shows that a gap exists, this one shows why it’s so hard to close: digital exclusion isn’t one problem with one fix, it’s several overlapping barriers that hit different groups of Australians in different ways. The Australian Digital Inclusion Index - co-produced by RMIT, Swinburne University of Technology, and Telstra - measures three separate dimensions of digital inclusion each year: whether people can get online at all (Access), whether they can afford to stay online (Affordability), and whether they have the skills and confidence to use the internet effectively (Digital Ability). The heatmap below lays out, group by group and dimension by dimension, exactly how far below the national average each group sits, using the report’s own published gap figures rather than a single blended score.

# SOURCE: ADII 2025 Report (RMIT, Swinburne, Telstra), pp.7-13.
# digitalinclusionindex.org.au/wp-content/uploads/2025/10/ADII-Report-2025_V6-Remediated.pdf
# Gap = points below the national average, reported per dimension in the
# report text. Blank/NA cells = not separately reported for that group.
gap_matrix <- tribble(
  ~group,                              ~dimension,        ~gap,
  "First Nations people",              "Total Index",     10.5,
  "First Nations people",              "Access",          12.9,
  "First Nations people",              "Affordability",   13.3,
  "First Nations people",              "Digital Ability",  5.3,
  "People with disability",            "Total Index",     11.4,
  "People with disability",            "Access",           6.3,
  "People with disability",            "Affordability",   13.1,
  "Public housing residents",          "Access",           9.3,
  "Public housing residents",          "Affordability",   26.0,
  "No secondary school completed",     "Affordability",   15.6,
  "No secondary school completed",     "Digital Ability", 19.1,
  "Capital city vs rest of Australia", "Total Index",       5.9,
  "Capital city vs rest of Australia", "Access",            4.5,
  "People experiencing unemployment",  "Affordability",   13.7
) %>%
  mutate(dimension = factor(dimension,
           levels = c("Access", "Affordability", "Digital Ability", "Total Index")))
p2 <- ggplot(gap_matrix, aes(x = dimension, y = fct_rev(group), fill = gap,
              text = paste0(group, "\n", dimension, ": ", gap, " points below national average"))) +
  geom_tile(colour = "white", linewidth = 1) +
  geom_text(aes(label = gap), colour = "white", size = 3.2, fontface = "bold") +
  scale_fill_gradient(low = "#FDECC8", high = "#6B4400", na.value = "grey92",
                       name = "Points\nbelow avg") +
  theme_minimal(base_size = 11) +
  theme(panel.grid = element_blank(), axis.text.x = element_text(angle = 20, hjust = 1)) +
  labs(
    subtitle = "Gap below national average, by group and dimension, 2025",
    x = NULL, y = NULL,
    caption = "Source: ADII 2025 Report, digitalinclusionindex.org.au. Blank cells = not separately reported."
  )

ggplotly(p2, tooltip = "text")

Read across any single row and a different story about the same disadvantage emerges depending on which dimension you look at. Public housing residents face their steepest barrier on cost - 26 points below the national average on Affordability alone, more than double the gap any other group faces on that dimension - while First Nations Australians are hit hardest on Access, at 12.9 points below average, reflecting infrastructure gaps in remote and regional communities that no amount of individual skill-building can fix on its own. These are, in large part, the same households sitting behind the falling literacy scores in the previous chart: a student cannot build digital literacy at school if their family cannot afford reliable internet at home, and no classroom intervention closes a gap that starts with the household budget.

Where you live still decides how included you are - and a standard map would have hidden that

Digital inclusion in Australia has a geography, and the shape of it isn’t the one a normal map would show you. Rather than plot each state and territory onto its true land area - where the sparsely populated Northern Territory and Western Australia would visually swallow the page purely because they are large, not because they represent where most digitally excluded Australians actually live - the chart below uses a tilegram: every state and territory gets one equal-sized tile, arranged to roughly preserve real compass directions (Northern Territory and Queensland up top, Western Australia to the left, Tasmania at the bottom) without letting geographic size distort the story. Each tile is coloured by whether that jurisdiction’s overall Digital Inclusion Index score sits above or below the 2025 national average of 73.6.

# SOURCE (data): ADII 2025 Report, p.11 (Figure 1). digitalinclusionindex.org.au
#
# CHART TYPE NOTE: built as a tilegram (equal-sized tiles, not true
# geography), following the Week 8 lecture method (geom_rect + reversed
# y-scale + Plotly hover), specifically to avoid two "Spatially
# Challenged" pitfalls flagged in Ch.7: (1) continuous colour scales, and
# (2) large-area prominence, where WA/NT/SA would visually dominate a
# true-geography choropleth despite representing a small share of the
# population this story is about. Tile positions approximate real
# relative geography (north/south/east/west) but are not to scale -
# that's the point of a tilegram.
#
# Fill is binned into "above/below the national average (73.6)" rather
# than a continuous gradient - this bin choice isn't arbitrary, it
# directly mirrors how the ADII 2025 report itself narrates the state
# results (NSW/VIC/WA "above the national average"; QLD/NT/SA "below").
adii_tile <- tribble(
  ~state, ~abbrev, ~index_score, ~col, ~row,
  "Northern Territory",           "NT",  72.5, 2, 1,
  "Queensland",                   "QLD", 72.9, 3, 1,
  "Western Australia",            "WA",  73.9, 1, 2,
  "South Australia",              "SA",  71.3, 2, 2,
  "New South Wales",              "NSW", 74.0, 3, 2,
  "Victoria",                     "VIC", 74.1, 2, 3,
  "Australian Capital Territory", "ACT", 79.5, 3, 3,
  "Tasmania",                     "TAS", 69.4, 2, 4
) %>%
  mutate(vs_average = if_else(index_score >= 73.6, "Above national average (73.6)",
                               "Below national average (73.6)"))
tile_size <- 0.46

p_map <- ggplot(adii_tile) +
  geom_rect(aes(xmin = col - tile_size, xmax = col + tile_size,
                ymin = row - tile_size, ymax = row + tile_size,
                fill = vs_average,
                text = paste0(state, ": ", index_score, " / 100\n", vs_average)),
            colour = "white", linewidth = 1.2) +
  geom_text(aes(x = col, y = row - 0.12, label = abbrev),
            colour = "white", fontface = "bold", size = 4) +
  geom_text(aes(x = col, y = row + 0.18, label = index_score),
            colour = "white", size = 3) +
  scale_fill_manual(values = c("Above national average (73.6)" = cb_purple,
                                "Below national average (73.6)" = cb_orange),
                     name = NULL) +
  scale_y_reverse() +
  coord_equal() +
  theme_void(base_size = 12) +
  theme(legend.position = "top", plot.margin = margin(t = 10, r = 10, b = 5, l = 10)) +
  labs(caption = "Source: ADII 2025 Report, digitalinclusionindex.org.au.\nTile layout is schematic, not to scale - avoids overweighting large, sparsely populated states.")

ggplotly(p_map, tooltip = "text") %>%
  layout(margin = list(t = 50, b = 60))

The pattern that survives this correction is a genuinely useful one: the divide runs less along a simple east-versus-west or north-versus-south line than along population density and socioeconomic composition. Tasmania records the lowest score in the country at 69.4, driven partly by the lowest median personal income of any state and an ageing population; South Australia and Queensland sit below average for similar structural reasons, the latter because its population is more spread out across regional centres with weaker Affordability and Access scores than the major capitals. The Australian Capital Territory, by contrast, posts the highest score nationally at 79.5 - a reminder that this map is really tracking income, age, and population concentration wearing a state-boundary disguise, not some innate geographic destiny.

Australian teachers have raced ahead on AI in the classroom - their training hasn’t caught up

The gap explored so far sits on the student side of the classroom. This chart turns to the adults responsible for closing it, using results from the OECD’s Teaching and Learning International Survey (TALIS), the world’s largest cross-country survey of teachers, fielded in 2024 across 55 education systems including Australia. The comparison below sits two different kinds of measurement side by side for the same teachers: how many actually used AI tools in their teaching over the past year, and - among those who didn’t - how many say the reason was a lack of the knowledge or skills to do so.

# SOURCE A: OECD Education GPS, TALIS 2024 Country Profile (Australia) -
# gpseducation.oecd.org - lower secondary AI use, rank 4/55.
# SOURCE B: ACER, Australian TALIS 2024 National Report (acer.org.au) -
# primary teacher AI use + non-user training gap.
talis_use <- tribble(
  ~teacher_level,   ~measure,                         ~pct,
  "Primary",        "Used AI in past year",            47,
  "Lower secondary", "Used AI in past year",           66,
  "Lower secondary", "OECD average (used AI)",         36,
  "Primary",        "Non-users citing lack of skills",  84,
  "Lower secondary","Non-users citing lack of skills",  75
)
p3 <- ggplot(talis_use, aes(x = teacher_level, y = pct, fill = measure,
              text = paste0(teacher_level, " - ", measure, ": ", pct, "%"))) +
  geom_col(position = position_dodge(width = 0.7), width = 0.6) +
  scale_fill_manual(values = c(
    "Used AI in past year" = cb_purple,
    "OECD average (used AI)" = tc_grey,
    "Non-users citing lack of skills" = cb_orange)) +
  scale_y_continuous(labels = label_percent(scale = 1)) +
  labs(
    subtitle = "% of teachers, by school level, TALIS 2024",
    x = NULL, y = NULL,
    caption = "Source: OECD TALIS 2024 (gpseducation.oecd.org); ACER Australian TALIS 2024 National Report (acer.org.au)."
  )

ggplotly(p3, tooltip = "text") %>%
  layout(legend = list(orientation = "h", y = 1.15), margin = list(t = 50))

Put the two measures next to each other and a genuinely counterintuitive picture appears: Australia isn’t lagging on AI adoption in schools at all - at 66% of lower secondary teachers having used AI in the past year, Australia ranks fourth out of all 55 education systems surveyed, nearly double the 36% OECD average. The shortfall is entirely on the follow-through. Of the Australian teachers who haven’t adopted AI, 75% of lower secondary and a striking 84% of primary teachers say it’s because they lack the skills or training to use it, not because they’re opposed to the technology or unconvinced of its value. Australia has, in effect, handed teachers a new tool faster than it has taught them how to use it - and it is students, not systems, who absorb the consequences of that gap first.

The further you climb the qualification ladder, the more the odds tilt in your favour

The first four visualisations trace where the AI-and-digital divide begins - unequal access, unequal skills, unequal teacher preparedness. This final chart follows that divide all the way to its economic endpoint: what actually happens to graduates once they leave the education system, according to the 2024 Graduate Outcomes Survey, Australia’s national survey of recent higher-education graduates, run by the Department of Education. Each bubble below represents one level of study - undergraduate, postgraduate coursework, or postgraduate research - positioned by two outcomes simultaneously: how likely graduates at that level were to be working full-time within four to six months of finishing, and how much they were earning if they were.

# SOURCE: Dept. of Education, 2024 GOS National Report - Figure 3
# (employment), Figure 5 (salary), Executive Summary (gender gap).
gos_bubble <- tribble(
  ~study_level,               ~ft_employment_pct, ~median_salary, ~gender_gap_pct,
  "Undergraduate",             74.0,               75000,          NA,
  "Postgraduate coursework",   88.1,               100000,         10.9,
  "Postgraduate research",     82.8,               104400,         NA
)
p_bubble <- ggplot(gos_bubble, aes(x = ft_employment_pct, y = median_salary,
              size = median_salary, colour = study_level,
              text = paste0(study_level,
                             "\nFull-time employment: ", ft_employment_pct, "%",
                             "\nMedian salary: $", comma(median_salary),
                             ifelse(is.na(gender_gap_pct), "",
                                    paste0("\nGender pay gap: ", gender_gap_pct, "%"))))) +
  geom_point(alpha = 0.85) +
  geom_text(aes(label = study_level), vjust = -2.3, size = 3.2, colour = cb_black, show.legend = FALSE) +
  scale_colour_manual(values = c("Undergraduate" = cb_purple,
                                  "Postgraduate coursework" = cb_orange,
                                  "Postgraduate research" = cb_gold)) +
  scale_size(range = c(10, 22), guide = "none") +
  scale_x_continuous(labels = label_percent(scale = 1), limits = c(65, 95)) +
  scale_y_continuous(labels = label_dollar(), expand = expansion(mult = c(0.15, 0.25))) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "none", panel.grid.minor = element_blank()) +
  labs(subtitle = "Full-time employment rate vs. median salary, 2024 domestic graduates",
       x = "Full-time employment rate", y = "Median annual salary",
       caption = "Source: Dept. of Education, 2024 GOS National Report, qilt.edu.au.")

ggplotly(p_bubble, tooltip = "text") %>%
  layout(margin = list(t = 50))

The distance between the three bubbles is the story: postgraduate coursework graduates were 14 percentage points more likely to be in full-time work than undergraduates, and earned a $25,000 higher median salary on top of that advantage - a gap that compounds rather than simply adds up, since the same graduates who struggle most to build digital literacy in the opening trend line, and who face the steepest access and affordability barriers in the digital disadvantage heatmap, are also the least likely to be positioned to pursue that extra qualification in the first place. Even at the postgraduate coursework level, where outcomes are strongest overall, a gender pay gap of 10.9% persists among graduates who completed the exact same kind of degree. The credential ladder rewards those who can climb it - but this report’s earlier charts make clear that the ability to start climbing was never distributed evenly to begin with.

Design notes

Two deliberate design choices, made in line with this course’s Chapter 3 (Visual Perception and Colour) and Chapter 7 (Spatial Data):

References

Australian Council for Educational Research. (2025). Australian TALIS 2024 national report. ACER. https://www.acer.org/au

Australian Curriculum, Assessment and Reporting Authority. (2026). NAP-ICT Literacy 2025 public report. ACARA. https://nap.edu.au/docs/default-source/nap-sample/nap-ictl-2025-public-report.pdf

Australian Curriculum, Assessment and Reporting Authority. (2023). NAP-ICT Literacy 2022 public report. ACARA. https://nap.edu.au

Department of Education. (2025). Short-term graduate outcomes in Australia: 2024 Graduate Outcomes Survey national report. Australian Government. https://www.qilt.edu.au

OECD. (2025). Results from TALIS 2024: Teaching for today’s world. OECD Publishing. https://www.oecd.org

Thomas, J., McCosker, A., Parkinson, S., Hegarty, K., Featherstone, D., Kennedy, J., Holcombe-James, I., Ormond-Parker, L., & Ganley, L. (2025). Measuring Australia’s digital divide: Australian Digital Inclusion Index 2025. RMIT University, Swinburne University of Technology & ACCAN. https://digitalinclusionindex.org.au