Segmenting Adult Web Users into Meaningful Age Categories

Segmenting Adult Web Users into Meaningful Age Categories Robert W. Bailey, Ph. D. Computer Psychology, Inc. bob@webusability. com 801 -2002

What are the 4 Most Useful Age Categories?

The Problem • Virtually every study separates adult participants differently, i. e. , designates different age segments • Without reading each individual study, practitioners do not know how old "old" is for each researcher • The goal is to have all researchers, who are doing work on ‘aging, ’ use the same age categories • Example – – Young: 20 -35 Middle-aged: 36 -55 Old: 56 -75 Old-old: 76 and over • What age segments are most useful to practitioners?

How Old is Old? • It is rumored that Otto von Bismark, Prime Minister of Prussia in the 1860’s introduced old age pensions • In preparation, he asked the mathematicians to determine the average of death • They found that it was “ 55” • He said, “We’ll pay pensions at 65”

How Old is Old? • One prominent medical doctor recently made these observations on aging • “Aging begins at 30” – Organs begin to lose their function – Increase of heart disease, diabetes, arthritis, etc. – Bones begin to become brittle • “Many practicing physicians now refer to the elderly as those 75 and older, and the ‘old-old’ as those 85 and older”

The Influence of Age and Experience on Data Entry Czaja and Sharit, 1997 • Past research - Young people perform reliably better than older people on speed-related tasks – Data entry – File modification – Inventory management • This study - Participants were 110 people who performed a data entry task for three days – Young - Mean of 29. 8 years – Middle - Mean of 49. 4 – Old - Mean of 66. 5 (Reliably less computer experience) • Results – Young and middle-aged users entered reliably more data than old users (p<. 001) – No age-related differences with errors

Age, Luminance and Print Legibility Charness and Dijkstra, 1999 • To survey homes, offices and public places – To determine existing ambient light levels – To assess whether ambient light levels in homes vary with the occupant’s age • To determine whether making changes to ambient light levels might improve the reading performance of older adults (intervention)

PAST RESEARCH • Study 1 - Older adults (aged 60 -83) – Read serif fonts (Roman) 6% faster than sans serif fonts – The best reading speeds were attained with 14 -point type • Study 2 - Older adults with an average of 75 – Read using 14 -point Times Roman and 9 -point Helvetica – 14 -point times was superior • Study 3 - Adults over age 50 were more strongly affected by low light levels than were people under 50 years of age

Study 1 • Participants were 98 Tallahassee residents – 31 young (20 -38, average 29) – 33 middle-aged (39 -58, average 47) – 34 older (over 58, average 69) • Performed five reading tasks • Results – The older group • Used reliably higher light levels • Read reliably slower than the younger groups – Adding lighting improved reading speed for all groups

Study 2 • Method – Visited 51 businesses – Measured the light level in work areas – Tested two people with reading tests • One over 40 • One under 40 • Results – Offices generally had adequate light levels – Only older users benefited from increasing the light level

Study 3 • Method – Visited 51 public places – Measured the light level in areas where people would read – Tested two people with reading tests • One over 50 • One under 50 • Results – The light levels in 71% of the locations were too low – Participants over age 50 read slower than those under 50

“Old” Defined • Past research – 60 -83 – Average of 75 – Over age 50 • These studies 1. 2. 3. 4. Mean of 66. 5 Over age 58 with an average of 69 Over age 40 Over age 50

Cognitive and Perceptual Training by Older and Younger Adults Mead and Fisk, 1997 • Investigated the type of information that should be presented during training – Young adults - Range of 18 -30 (mean = 20) – Older adults - Range of 64 -80 (mean = 69. 9) • The groups showed no reliable differences on – Simple reaction time tests – Corrected vision tests

Young vs. Older Users • Young adults – Were more likely to have used an ATM (p<. 0001) – Used computers more often (p<. 0001) – Had higher scores on • • Perceptual speed (p<. 0001) Reading rate (p<. 05) Reading comprehension (p<. 0001) Working memory capacity (p<. 0001) – Had faster choice reaction times (p<. 0001) • Older adults – Were better educated (p<. 05) – Had higher vocabulary scores (p<. 05)

How Old are Your Participants? An Investigation of Age Classifications Timothy A. Nichols, Wendy A. Rogers, Arthur D. Fisk, and Lacy D. West Georgia Institute of Technology Proceedings of the Human Factors and Ergonomics Society 45 th Annual Meeting 2001

Introduction • Designers should try to account for age-related differences in their user populations • Gathered reported age data from all articles from two journals – Human Factors Journal: 1998 -2000 – Psychology & Aging: 1995 -1999 • Attempted to determine how researchers segmented their participants by age

Human Factors Journal • Human Factors Journal reported 131 empirical articles – 49 (37%) provided no age data at all – 64 (51%) supplied some information – 18 (14%) listed a mean, standard deviation and age ranges • Psychology & Aging reported 202 empirical articles

Results Classification Older Middle-aged Young HF 58 -76 40 -59 19 -35 P&A 62 -82 41 -57 19 -30

Longitudinal vs. Cross-Sectional Studies

Chronological Age • Cannot “cause” anything • Can help in defining the probability of occurrence of certain events

Age and Experience Relationships Sri Kurniawan, Jason Allaire and Darin Ellis, 1999 • Examined the relationships among age, web experience and web ability • Participants were 600 older adults (average of 44. 3 years) • About 45% of the variance in Web ability was explained by the user’s age and experience – Web experience - 28% of the variance – Age - 9% of the variance – Shared age and experience - 8% of the variance

Longitudinal vs. Cross-Sectional Studies • Longitudinal - Compare the same individuals over time (historical effects) • Cross-sectional – Individuals are compared within their age groups – May belong to different age cohorts – May have had different life experiences • The findings from the two types of studies do not always agree

Cross-Sectional vs. Longitudinal

Studying the Effects Aging • Longitudinal – Measures the changes in one group of people over time – Usually considered superior to cross-sectional – Can be confounded by • • Selection bias Selective attrition Retest familiarization ‘Historical’ effects (see ‘world record’ times) • Cross-sectional – Evaluates for differences across the different age groups – Can be confounded by • • Older adults being more cautious (work slower) Major educational and experience differences Slowing of the central nervous system over a certain age Some differences can be the result of testing only survivors (those who have not yet died)

World Record Times personal. rdg. ac. uk 1912 1920 1924 1928 1932 1936 1948 1952 1956 1960 1964 10. 5 seconds 10. 5 10. 2 10. 1 10. 0 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 9. 95 seconds 9. 95 9. 93 9. 86 9. 84 9. 79 9. 78

Abilities and Age • Data from longitudinal studies will better measure age changes for those in – Good health, and – Stimulating environments • Data from cross-sectional studies tend to over estimate loss of most abilities • Cohort effects (e. g. , differences in the amount of education) usually accounts for more variance than age-related factors

Six Ages of Humans Pirow, 1994 • Birth • Starting age - The earliest age at which a measured activity can take place • Competence - The age at which a person has acquired the skill to perform well • Optimal - The age at which the person will perform optimally at the task • Initial decrease - The age at which the performance will start to decrease linearly • Rapid decrease - The age after which the performance will decrease at an increasing rate

Running Example Starting Competence Optimal Initial decrease Rapid decrease Female Male 2 years 9 10 22 24 24 29 59 66

Correlation of Track and Field Performance with Chronological Aging Fung and Ha, 1994 400 meters 1500 meters 200 meters 800 meters 5000 meters 100 meters High jump Discus Shot put Javelin Correlation Female Male. 98. 97. 96. 95. 97. 94. 98. 94. 96. 92. 94. 88. 91. 83. 78. 81. 79. 74. 94

General Decline in Older Adults • • Sensitivity of most sensory organs Attention capacities Working memory Speed of motor performance

Abilities and Age Woolf, 1998 • Reliable decrements can not be found for all abilities for all persons (until very late in life) • Decline is most evident where ‘speed of response’ is involved • Declines will be evident in most abilities – For those in their 50 s and 60 s who live in deprived environments, and – For individuals of any age who have severe central nervous system disease (e. g. , Alzheimer’s)

Common Age-Related Changes in Vision • Decreased sharpness of vision (visual acuity) • Decreased ability to focus on near objects • Decreased ability to focus on objects at varying distances (visual accommodation) • Decreased ability to discriminate between certain color intensities – Especially in the blue-green end of the color spectrum – The "yellowing" of the lens with age makes blues and greens appear "washed out" or faded • Decreased ability to perceive or judge depth • Decreased ability to focus in low light levels • Slow responsiveness to changes in light levels (dark to light, and light to dark) • Increased sensitivity to glare • Decreased ability to accurately judge distances • Increased need for light needed to see objects clearly

Age-Related Changes in Vision

Comfortable Listening Levels Coren, 1994 • The number of people who have difficulty hearing and understanding voices increases with age – General conversations – Voices on • • The phone Television Radio Computer • Procedure – Used 799 subjects, ranging in age from 17 to 92 – Each • Listened to a running speech signal • Identified the level preferred for listening

Results • The average `most comfortable listening level' for all participants was 63. 4 d. B • They found – No differences between left and right ears – No differences between male and female • Before the age of 40, the most comfortable listening level increased about 1/3 d. B per year • After the age of 65, the most comfortable listening level increased about 1/2 d. B per year

Hearing Comfort Level by Age

Primary Mental Abilities Schaie, 1958

Primary Mental Abilities Shaie, 1972

Reaction Time Fozard, 1990 • Shortens from infancy into the late 20 s • Increases slowly until the 50 s and 60 s • Lengthens faster as a person gets into the 70 s and beyond • Becomes more variable with age • When troubled by a distraction, older people tend to devote their exclusive attention to one stimulus and ignore another (attention)

Normal Distributions by Age

Normal Distributions by Age Slower Means

Normal Distributions by Age Slower Means and More Variability

Longitudinal Analysis of Age. Related Slowing Fozard, et. al. , 1990 • The Baltimore Longitudinal Study of Aging has been gathering data since 1959 – 1300 adults from 20 to 96 years of age – Continually evaluated using different measures • Biographical • Physiological • Psychological • One measure is reaction times – Simple – Responded to both high (1000 Hz) and low (250 Hz) tones presented for 3 seconds at 62 d. BA – Disjunctive (choice) – Responded only to high tones

Results • Reaction time increases with age – Constant rate of slowing over the adult life span – appears linear – Slows from 10 -20 milliseconds per decade (12 milliseconds per year) – Men remain reliably faster than women • There seems to be a general slowing of central nervous system functions with aging

Aging and Computer‑based Task Performance Sharit and Czaja, 1994 • Of particular interest are age‑related changes in information processing abilities, including the – Senses – Cognition processors – Responders • There seems to be a general overall slowing in cognitive tasks • The hypothesized `slowing factor' for cognitive tasks is 1: 1. 6 (young vs. old)

Cognitive Abilities and Job Performance • There is little evidence that job performance declines with age • Age alone is not a significant predictor of performance in most actual work activities • Age effects are – Smaller for tasks where knowledge is an important aspect of the task – Larger for tasks where successful performance is primarily dependent on speed

Performance on Choice Reaction Time and Typing Tasks

Aging and Errors Rabbitt, 1990 • Used a two-choice reaction time task • Four age groups – – 19 -30 50 -59 60 -69 70 -79 • Conditions – No response to errors – Corrected each detected error – Signaled that an error was made (no correction)

Results • All age groups – Made the same percentage of errors – Were equally proficient at ‘automatic’ error detection – Underestimated the number of errors made (after the test) • The 70 -79 group ‘signaled’ reliably fewer errors • The ability to remember errors after the test declined with age – beginning at age 50

100 Meters Record by Age world-masters-athletics. org

Mile Run Records by Age home. hetnet. nl

High Jump Records by Age world-masters-athletics. org

Shot Put Records by Age world-masters-athletics. org

Common Web-based Tasks • • Typing Mousing Linking Paging Using widgets Scrolling Reading etc.

Comparing Age Groups Koyani, Bailey, Ahmadi, Changkit and Harley, 2002 Ages 20 -30 with Ages 71 -80

Comparing Age Groups 20 -30, ‘ 61 -70’, 71 -80

Interventions • • Eyeglasses, contact lens, hearing aids Recall vs. recognition memory Length and type of training TFT vs. CRT screens Intensity (loudness) of auditory signals Shape of the cursor Time of day Accessibility features

Mechanisms of Human Aging

Cognitive Correlates of Human Brain Aging Coffey, et al. , 2001 • Collected MRI data from 320 volunteers (ages 66 -90) • Compared the results with performance on – Attention – Information processing speed, and – Memory • The findings suggest a relationship between age -related changes in brain structure and declines in attention, psychomotor speed and memory

Age-Related Gray and White Matter Changes – Longitudinal • • Resnick, et al. , 2003 Conducted MRIs on 92 non-demented older adults aged 59 -85 Used the baseline, 2 year and 4 year follow-ups in the BLSA Found reliable age decreases in both gray and white matter There seemed to be slower rates of brain atrophy in individuals who remained medically and cognitively healthy

Brain’s Gray and White Matter

Determining Gray vs. White Matter

Age-Related Gray and White Matter Changes – Cross-Sectional Ge, et al. , 2002 • 54 healthy volunteers aged 20 to 86 were given MRIs • Findings – The percent of gray matter and white matter were reliably less in older (over age 50) adults – The percent of gray matter decreased linearly with age – beginning with the youngest participants – There was no difference between sexes

Gray Matter

Does Loss of Brain Tissue Accelerate as People Get Older? sciencedaily. com, 1998 • Divided patients into three age groups: – Young-old: 65 -74 years old – Middle-old: 75 -84 years old – Oldest-old: 85 -95 years old • Measured the changes in brain volume with magnetic resonance imaging (MRI) scans • The loss of tissue among patients was a constant 1% or less per year • Dementia is related to a more rapid brain tissue loss

Is Cognitive Decline Normal? Haan, et al. , 1999 • Tracked changes in cardiovascular health, diabetes and cognitive function over a 7 -year period • The people were all 65 or over when recruited • 70% of the individuals showed no significant decline in cognitive function (Modified Mini-Mental State Exam) • The greatest loss of cognitive ability occurred in people who had – High levels of atherosclerosis or diabetes, and – The apolipoprotein E 4 gene (Apo. E 4) • They were 8 times more likely to show a decline in cognitive function

Chromosomes • Humans have 23 chromosomes • Twenty-two are numbered in order of size – Largest (number 1) – Smallest (number 22)

Chromosomes

Genes • Each chromosome contains genes • Genes are stretches of (deoxyribonucleic acid) DNA that comprise the recipes for proteins

Damaged Genes = Cognitive Decline Lu, et al. , 2004 • Damaged genes can start in the late 30 s and early 40 s in some individuals (i. e. , functioning at a reduced level) • Evaluated patterns of gene expression in postmortem samples – Collected from 30 individuals – Ranged in age from 26 to 106 • Found two groups of genes with altered expression levels – Those related to learning and memory – Those related to gene repair mechanisms • Conclusion: DNA damage may reduce the expression of certain vulnerable genes involved in learning, memory and neuronal survival

Proposed Age Categories • • Old-old: 75 and older Older: 60 -74 Middle-aged: 40 -59 Young: 18 -39

Possibly More Important • Overall level of cognitive activity • Severe nervous system diseases – Alzheimer’s – Parkinson’s • Circulation-related diseases – Atherosclerosis – Diabetes • Certain medications • Deprived environment • Seriously hampered senses – – Cataracts Glaucoma Macular degeneration Diabetic retinopathy • Defective genes (DNA)