Act Early: Holme Wood Introduction

n 10 January 2020, 124 people came together to talk about Holme OWood — a single locality in the District. The attendees comprised front-line practitioners from community organisations, public service providers, local elected members and policy makers, alongside some of the world’s leading data scientists via invitation from The Alan Turing Institute, the UK’s national centre for Artificial Intelligence and Data Science. The purpose of the meeting was to address two simple questions: “Can data science improve our collective understanding of a place like Holme Wood?” “Can we use data to help the community strengthen and grow by ‘acting together’?”

In short, there was a positive answer (‘yes!’) to both questions. Our attendees’ observations and questions generated five action research projects, which we will be taking forward together. Crucially, the lines of enquiry for each project cut across traditional public service boundaries, reflecting the lived experience of people living and working in Holme Wood. Annex A provides more detail on the process followed on the day.

2 The five projects are:

‡ Exclusions, children not in school, and crime

‡ Healthy choices: healthy lives1

‡ The impact of poor mental health on individuals, the community and services2

‡ Pride, aspirations, role models and careers

‡ Adverse Childhood Experiences

All projects will include an economic line of enquiry, considering the costs of current services and of issues playing out unaddressed — the costs to individuals, the community and services, plus the ‘opportunity cost’ — the potential positive financial effect of improving the situation and the cost of inaction.

Annex B provides more detail on each project and the initial lines of enquiry proposed by our groups.

Annex C outlines the initial ideas and questions discussed on the day by data scientists, on analytics, AI techniques and the visualisation of information.

How we will work? Our principles for delivery

Holme Wood has been a focus for concern, action and investment for many years. Numerous projects, including large scale investments, appear to have failed to achieve

1 Includes dental care – impact, access, choices and food (including food impact, access and choices) 2 Includes questions raised on the impact of mental health on other public services, and the topic of mental health and isolation 3 0

4 their goals. A common complaint is of ‘top • For Bradford, and beyond, this project will test, down’ interventions, where residents and local model and share a process that brings cutting organisations are the recipients of support, advice edge data science into the conversation about of investment, against goals and targets set by public services and public policy, to improve others, and often proved to be unattainable. outcomes for all. Our vision is of a process (a We are committed to making Act Early: Holme “place-based approach” to working) that can Wo o d different. be replicated across Bradford, adopted by other communities, and help those involved in policy First, our projects will be open to, and shaped by, the making at national level, design programmes people who live and work in Holme Wood. As this that can adapt to local circumstances. We further report confirms, our first phase of activity will see us anticipate that Act Early: Holme Wood will allow reach out to the community, testing the issues identified the scientific community to sharpen its techniques by practitioners and others against residents’ perceptions and data science tools. and experiences of what matters, what works, and what needs to change. This will take place alongside a similar testing of the issues by our data scientists. Why Bradford? We will be ready to adapt our action research projects to take account of what both of these important groups tell us. As the projects progress, our data scientists The desire to apply data science to public service with expertise in visualisation will produce accessible design is not unique to Bradford or Act Early. But, models that can be manipulated and challenged as Bradford is uniquely placed to host these project easily by residents as by practitioners, policy makers because it has the following resources: and the scientists themselves. Projects will encourage participation by residents, taking part and supported • The Born in Bradford project, one of the world’s to act as citizen scientists. Projects will be governed by largest longitudinal birth cohort studies tracking the groups including members of the community. lives of 13,818 children (trailblazing data linkage via Connected Bradford3) Second, we will be realistic about what these • Regional expertise in the Digital Triangle projects can achieve for Holme Wood. Act Early (comprising AI, Data Analytics and Immersive is about applying science to improve public health Visualisation Technologies) outcomes (including education, health and economic wellbeing). Act Early funding is available to drive the • Diverse demographics (ensuring ethnic and science. It cannot be used to pay for additional front socioeconomic diversity) line services. • A national spotlight on Bradford (Bradford is benefiting from additional support through its But, this does not mean we lack ambition. Opportunity Area (Department for Education) has two important goals: Act Early: Holme Wood and Integration Area (Ministry for Housing, Communities and Local Government)) • For Holme Wood, it will propose to Bradford’s decision makers and service planners, ways to more • Support from all key decision makers and the effectively align and target services, with a focus on community early intervention. It will suggest ways to empower communities to make informed choices, take more Bradford also has a proud history of using science to ownership of issues, and improve access to services. improve public health and education. In 1907, Green Local decision makers are already represented on Lane Primary School began offering free school meals to our Act Early: Holme Wood ‘leadership group’, its pupils. This was more than an innovation in health, and have committed to participate fully in the social and economic policy. It also represented an action research projects, and to take seriously innovation in and commitment to applying data science recommendations for change. to improve children’s outcomes, because the school also

3 Connected Bradford builds on the data linkage undertaken by Born in Bradford and connects all of the routinely held data on children’s health and education across the Bradford district (going back 10 years in time) 5 recorded changes in their pupils the action research projects, community data (e.g. through weight, from the introduction of establishing online networking surveying the community the free meals. The data showed sites for each project, allowing we can explore whether children gaining weight when the for exchange of ideas and community responses indicate school was open and losing weight information. underreporting). when it was closed; education and • We will provide the relevant • We will use this phase of health interacting. Ultimately, the anonymised local and national enquiry to produce models that Green Lane study helped to inform data sets4 to the scientists, allow scientists, practitioners national policy on school meals. asking them identify patterns and residents to visualise and and relationships relevant to manipulate the models and test the identified issues, including different predictions. Our plan for delivery. predictive indicators. This will • We will widen our include testing the validity conversation with front line Each of the projects will follow a of perceived predictive, staff, service planners and similar path: causal and remedial factors national policy makers; at the suggested by front line same time engaging directly practitioners, alongside Phase 1: with Holme Wood residents interactive data visualisation and community organisations. tools to communicate these relationships. • This process will help each • We will ask data scientists from project to develop a richer the Turing Institute, the Leeds • Publically held datasets will understanding of the issues Institute of Data Analytics (e.g. be tested against datasets and the lived experiences of the ), generated by the community residents (and those supporting and the Regional Universities to identify if disparities exist them), as well as develop to link to one or more of between the official and

4 We have been working with local partners to secure the relevant data sets, since autumn 2019, including ensuring robust data protection measures 6 Promoting the action research People living Data Practitioners Service projects; inviting and working planners and people to join one scientists and working in researchers in Holme policy makers or more groups Wood Holme Wood

Action Research Project Group

Community and Data Science Activity Wider Consultation

Within each action research Linking and analysing Training in ‘citizen group, data datasets that follow science’ available to scientists take and refine the initial residents, including forward analysis lines of enquiry under work with schools and development each identified Action of models; in Research Project parallel, Testing and refining community and lines of enquiry, with other partners ‘residents’, discuss lines of practitioners and enquiry; data service planners/policy scientists makers — including, additionally where possible, improve the national and regional quality, depth policy makers and breadth of data through local collection; Development of models and development and creating New ‘community of citizen science. visualisations that generated datasets’ ensure accessibility and created to fill gaps and usability provide context

Groups come back together to undertake further actions Action Research Project Group

Sharing, testing and refining models and visualisation tools Using models to refine lines of enquiry Developing proposals for action (changes to frontline practice etc.) as well as evaluation methodology, communications, governance etc.

Working in new Plan ways, together, as proposed by the Action Report Deliver Research Groups Evaluate 7 new ideas to test or resolve behaviour), and establishing: Phase 4: identified issues. It will also (i) clear project timelines; (ii) allow us to build the team robust regular monitoring, around each project. feedback and evaluation The fourth and final phase will processes that allow for rapid • We will work through conclude the research stage of iteration. established networks and the project — delivering an partner organisations — • Learning from the process evaluation report containing such as Bradford’s Stronger followed to this point, we recommendations for change to Communities partnership — to will work with local partners local decision makers — a new reach staff and run events in to identify at least one and local plan — signed off by the the community. potentially two more ‘places’ in community, data scientists and Bradford, where we can start to staff. • We will launch an offer of apply a similar methodology. practical training for residents, We will also deliver a process including work with students review of the methodology, with in schools, focusing on science Phase 3: its own recommendations for and digital literacy. This will improving and replicating the link to our broader Act Early approach. We expect this to result and Born in Bradford goal of In phase three, our projects in an academic publication that promoting citizen science, and will be delivered through to provides a novel contribution to strengthening the plans around completion. End dates will vary the scientific literature on how to making Bradford ‘the of from project to project. implement place‑based working Research’ (a trailblazing site through data analytics. for scientific working across communities).

Phase 2:

• We will bring the scientists, policy makers, practitioners and residents back together, in the 5 action research groupings. This will allow each group to share and explore the findings from the data analyses and local conversations.

• These meetings will also test the usefulness of different visualisation techniques in helping people engage with and manipulate the data.

• Each group will define and set up their action research project, setting a hypothesis for what will change in front line practice (or other areas of service planning, monitoring, targeting, or individual

8 A n n e x A : Summary of the process we followed on 10 January

Attendees were split into six were asked to respond to the priorities brought through ‘specialist’ groups: from Session 1 by the front line group. Session 2 asked participants to develop a first sketch of action research projects matching both the front line groups’ priorities • Two groups of front line staff working in Holme and appearing likely to support robust data analytics. Wood, including schools, police, housing, education, community organisations etc. Over lunch, the organising team collated • A group of senior officers and planners responsible the sketches produced by each of the six for the design of local services (from here onwards multi‑disciplinary teams, producing a common described as the ‘Planners group’) list of eight projects based on the most common themes to be taken forward for discussion in • A group of data scientists specialising in Artificial Session 3. Other project ideas were recorded for Intelligence techniques (e.g. Machine Learning) future action. • A group of data scientists specialising in ‘Visualisation’ of data (including immersive In Session 3: technologies and associated interfaces that allow Every delegate returned to their specialist group. intuitive interaction with datasets) Each group was asked to comment on some of the proposed projects, and develop ideas for moving the • A group of data scientists specialising in Urban eight projects into delivery. The data scientists were Analytics and data modelling asked to focus on methodology. The policy group and front line groups considered how local services could In Session 1: support projects and how the conversation could be We asked the ‘Holme Wood’ and the ‘Planners’ group broadened to engage the community. In practice, the to identify the priorities for Holme Wood within the time available meant most groups focused on two or specialist groups detailed above. The data scientists three projects in detail. shared — in their individual groups — ideas for using data science to better understand places and their In Session 4: social issues. The multi‑disciplinary groups (formed in the second session), sought to refine the action research proposals: In Session 2: defining the lines of enquiry that could direct data We created six multi‑disciplinary groups, each scientists, practitioner consultations, and discussions including specialists from all three data science groups, with the community. the planning group, and representatives from the Holme Wood ‘front line’ group. Data scientists and planners

9 Annex B: Detail o n t h e p r o j e c t s and proposed lines of enquiry

10 Project 1: Lines of enquiry: asked specifically to examine the relationship between Exclusions, defining the issue in exclusions and crime. children not in Holme Wood • The teams also reported a school, and crime need to examine the influence • What are the characteristics of of the following factors on children excluded from school school exclusions: in Holme Wood? Context — known and • Poor school attendance • What does the literature • Bullying perceived issues identify as the predictive indicators for children likely to • Children vulnerable In Bradford, 10.5% of state-funded cause concern to the police? because of issues at Secondary School students (4034 primary-secondary • What are the characteristics of students) were excluded for part of transition Holme Wood children causing the school year, with almost 5.9% concern to the police? • Prior speech and language of students being excluded multiple issues (and whether these times in the same year5. Persistent • Does crime follow exclusion or needs have been met) disruptive behaviour, and verbal the other way around — what or physical abuse towards other affects the sequencing? • Unsupported emotional students or adults accounted for and mental health issues • What are the preliminary most of these exclusions. Excluded behaviours for exclusions, • A lack of confidence children have been shown to have dropping out of school, and for in academic ability — poorer outcomes including mental criminal behaviour (attendance, potentially linked to a health and social exclusion. anti‑social behaviour etc.)? point in time such as mock exam results Within the Holme Wood • What are the schools serving workshop, concerns were raised Holme Wood’s policies on • Poor prior attainment regarding the outcomes for exclusion, are they consistent (any stage) these children, as well as the across the area; are their rates • Contact with gangs link between exclusions and of and reasons for exclusion later involvement in crime (both consistent with similar schools • Parental behaviour, in as victims and perpetrators). elsewhere (similar in cohort particular law breaking There is some evidence from and policy)? • Access to Early Help other longitudinal studies to link • What concerns do children exclusions and later criminal • Access to social workers cause to the police in Holme convictions, but whether Wood? In particular: • Access to out of school and there is a causal relationship holiday activities is unclear and could be tested • Which offences? here. There is evidence from • Who else is involved? the scientific literature that key Lines of enquiry: the factors associated with exclusion • Do we know anything economic case for include male gender, ethnicity, about unreported crime? lower socio‑economic status, action in Holme Wood • What are the interactions parental factors, mental health between Holme Wood schools and behavioural difficulties, • Can we identify the full costs and the police? (the group social communication difficulties, of implementing the current noted every primary school language difficulties, antisocial system, support and action by has a police presence, although activities, bullying/being bullied, schools, including referrals to criminal activity may be more low attainment, and special PRU and Alternative Provision prevalent in secondary). Our educational needs. (AP) and delivery of police practitioners and policy makers services, the impact of crime

5 Department for Education (2019) ‘Permanent and fixed period exclusions in : 2017 to 2018’ 11 on the community, as well as the opportunity costs • Access to social workers for the children themselves — i.e. hours lost to • Letters from family members in prison learning, impact on attainment and future earnings?

• Can we predict scope for savings in Holme Wood, and extrapolate beyond this area (district, national)? Project 2: Healthy choices, healthy lives Suggested data sources

• Attendance data, exclusions data, managed moves Context — known and and unexplained pupil exits perceived issues • Police arrests and incident reports

• Home schooled children Life expectancy is 9.1 years lower for men and 7.8 years lower for women in the most deprived areas of Bradford • Children missing from education when compared to the least deprived areas. In Year 5, 24.3% (1,705) of children are classified as obese, worse Potential responses than the average for England. The areas where Bradford is worse than National average figures includes: the rate • What does the literature suggest are the protective for alcohol-specific hospital admissions among those and preventative factors against exclusion, and how under 18 years; smoking during pregnancy; the rate many of these are routinely, effectively adopted in for alcohol-related harm hospital admissions; the rate Holme Wood? for self-harm hospital admissions; levels of smoking prevalence in adults; physical activity levels in adults; • What does the literature suggest are the protective rates of violent crime (hospital admissions for violence); and preventative factors against criminal activity, mortality rates from cardiovascular diseases and cancer. and how many of these are routinely, effectively delivered in Holme Wood? Concerns were raised by the community at the • Who already works with children most at risk of workshop regarding the lack of access to affordable, exclusion and/or crime, in Holme Wood? What healthy food. There appear to be a huge number of effect can we see on exclusions and crime (or takeaways and convenience stores on the doorstop, identified predictive/preventative factors)? with supermarkets and healthier options only available out‑of‑area. The consequences of poor food choices Possible implementation such as poor teeth and obesity were also highlighted. challenges and opportunities Lines of enquiry: defining the issue • There may well be a lack of data on the most in Holme Wood vulnerable children — those excluded but not yet formally engaged by the police • What do the data show regarding healthy/ unhealthy food availability across the area? • There is a risk that the number of children, once divided into cohorts (for characteristics, type of • Can we overlay obesity rates on takeaway criminal activity, reasons for exclusions etc) could density? Number of outlets per capita? be too small to support machine learning techniques • Can we map use of take-aways, including • A need to be careful about the language of ‘crime’ delivery services (Just-Eat etc)?

• An opportunity to evaluate, learn from, work with: • Proximity of takeaways to patterns of obesity?

• The Valley Project • Map dental issues against food availability

• PRUs and other non-standard school settings • Map dental service access for the area

12 • Can we overlay green spaces and leisure/sports?

• Can we overlay travel patterns and modes?

• Can we track people’s lifestyle and diet — what are people cooking and buying in Holme Wood?

• What are children and adults in Holme Wood being taught about healthy nutrition and cooking?

• What are the schools offering on school food choices, packed lunch policy, and on pupils accessing take away food? What role does a (healthy) breakfast club play?

• What impact does poor nutrition have on school attendance?

• What data are available from food banks about access by families, and the food ‘choices’ families make?

• What could supermarkets and other providers share on purchasing choices?

• Can we see a relationship in Holme Wood between poor food choices and adult literacy?

• Can we identify a location(s) for comparison outside Holme Wood?

• What is the council’s policy on planning etc. for takeaways and supermarkets — is this comparable to other areas in and beyond Bradford, and is the effect of planning policy comparable?

13 Lines of enquiry: the economic case changing behaviour, and how many of these are for action routinely and effectively adopted in Holme Wood? • What organisations and facilities are in place to • Costs of poor health to individuals and services, deliver such approaches? including: impact in the classroom for young • Who already works with the families who we people; consequence for the workplace or want to support (on any project)? accessing work because of ill health; subsequent pressures on the health service Potential implementation challenges • Can we estimate the financial pressures on families’ food and cooking budgets as well and opportunities as time constraints and any other resource barriers – access to kitchen equipment? • An opportunity to evaluate, learn from, work with: Include price of cooking against takeaways • Manchester City’s ‘City Cooks’ programme and ready meals • Rose Dunlop from ‘Living Well’ • What role could supermarkets play in shaping and offering support and better choices? • A local project [ODI — Scores on the Doors] creating open source database from hygiene • Would adapting transport options help with certificates for a number of fast food outlets, changing behaviour? and comparison to other communities — • Can we model the effect of introducing new shops? running to 2021

• Food banks (St Christopher’s, St Columbus’s, Suggested data sources Salvation Army, TFD-library)

• HENRY — healthy families group • Public health data programmes in Better Start Bradford areas • Typical consumption data in Holme Wood (www.henry.org.uk/betterstartbradford/ findaprogramme) • The price difference between fast food and home cooked food • The JU:MP project and its evaluation (www.activebradford.com) • The locations and quantity of fast food outlets

• The locations and quantity of ‘good food’ outlets • Take away/delivery data Project 3: The impact of poor • School absence data, particularly where this is mental health directly caused by nutritional issues, such as poor dental health (and dental appointments data)

Potential responses Context — known and perceived issues • What does the literature suggest are the factors affecting: Bradford District has a higher prevalence of children’s • People’s food choices? mental health issues than the average for England. • People’s willingness and ability to travel to buy Bradford has more hospital admissions as a result better food? of self-harm; a higher prevalence of children with conduct disorders; a higher prevalence of children with • How does this compare to what we know about emotional problems; a higher percentage of primary and Holme Wood residents? secondary school children identified with an SEN where • What does the literature suggest are the most the primary need is social, emotional and mental health. effective approaches to educating individuals and In summary, children within Bradford are, on average,

14 15 at increased risk for mental health disorders than other • Children and young people children in England. • Isolation is a key issue:

At the workshop, the community • What leads to isolation? felt: • Link to income? • Impact on children? • There was low funding and not enough resource • Link to lack of services/facilities? allocated for the police to support mental health (much of their resource went to dealing with 999 • What interventions would work to reduce isolation? calls from people in crisis) • New building in Holme Wood — considering • There was a serious need for more resources and putting professionals in but would lose community better allocation of systems to support mental health space. Can we simulate what the effect would be? problems Would need right data and would require a lot of extra data collection. Use of community centre for • Within the area, there was little counselling available services relating to isolation? Or spread out services? and the fact that CAHMS is located out of the area was a significant barrier to accessing services

• Loneliness, social isolation and resilience were key Lines of enquiry: the economic case factors. If people are socially isolated, they call the for action police because they have no idea who to turn to

• Consistent long-term solutions are needed • Could the data scientists help with making tools available to help with evaluation of interventions • The focus should be on early intervention rather implemented in the community with a focus on than crisis management cost (of services)? This would help identify where additional funding or policy change is required

Lines of enquiry: defining the issue • Focus on early prevention is key, but need to in Holme Wood demonstrate interventions are cost effective — is there a way of modelling this to assist policy makers/funders? • Can we map out the services and provisions currently available to be used as a resource for police etc. who can direct people (and for the Suggested data sources community to use)?

• Could we present this visually (e.g. social spaces)? • Clinical referrals

• Show which services are free and which cost • Police call-out data money to attend • A&E visits for self-harm • Use map to identify what services or assets are • Social space/asset data missing in the area • Health data (physical and mental) • Speak to people about what they would want

• Can we map where people suffering from mental ill health come into contact with A&E, police, Potential responses CAHMS etc. to understand need in the area? • Map out the services and provisions currently • Key groups to focus on: available in Holme Wood and estimate demand • Young men • Look at the evidence as to what works in reducing • Single and new mothers social isolation, and which services are missing

16 • Model the effect of introducing • Potential contacts/partners: a new/enhanced service • Holme Wood walking focusing on early prevention — group (NHS sponsored) what would be the impact for the individual and for different • Primary care networks services? • GPs

• Holme Wood library Potential • St Christopher’s good implementation neighbour (www. challenges and goodneighbour.co.uk/about) opportunities

• The scale is so big, it can seem overwhelming

• Issues are everywhere

• Issues seem too big to fix

• There are too many families with ‘needs’

• We don’t even know all of the families needing help (only a subset of people in need are known to services)

• There is a data gap with social isolation

• Mental health is linked to deprivation and poverty, which is a symptom of lots of other problems

• Difficulties with modelling (e.g. if free up time for police by reducing callouts, then have more time in other service provision) — need to show far reaching consequences

• These are sensitive issues — need to distinguish between situations where computer modelling is appropriate and others where need to use human judgement

17 Project 4: Pride, aspirations, There is a lack (or lack of awareness) of good quality jobs and stable employment, especially within the area. role models and careers Lines of enquiry: defining the issue Context — known and in Holme Wood perceived issues Tailoring the curriculum to match children needs and aspirations and ensure visible positive role models. People are ‘stuck’ — people can’t make the positive decision to live in Holme Wood (they are either there What is the link between education and employment? because they are given a , can’t get out or Trying to understand why students drop out is don’t know they can get out — it’s a negative choice or important. Addressing the re-inclusion of excluded no choice) — successful people tend to leave (but with people (expelled students don’t go back to school). many notable exceptions). Examine the relationship with parenting and the social circle (social context matters). This theme is very important to the police — they perceive that a lack of, or negative, role models/ What affects the ability to make good life decisions and aspirations makes children vulnerable and leads to can we model this? bad choices (more easily influenced and likely to get involved in crime). • Adverse Childhood Experience (ACE)

• Parenting style • For example, children say they want to become a drug dealer (have money, fancy car) when asked in • Poverty school about future job plans • Lack of aspiration • A lot of people work hard but stay quiet and • Income management (debt, smoking expenditure) are unseen as they don’t engage much with the community (unlike the clearly present drug dealer • Stress in their fancy car) • Social management • Lots of the positive role models end up leaving • Lack of choice Holme Wood whereas the negative ones stay • Influencers of young people Acceptance and lack of hope are a big issue: • Making decisions at a young age • Hard to change if people don’t care (e.g. rubbish in garden). The attitude is ‘we are currently coping, Lines of enquiry: the economic case we think we are fine — we don’t want to change’ for action • People may not even know that there is a choice (the problems are so embedded, and people are Could the data scientists make tools available to not empowered) help with evaluation of interventions [already • There is a mind-set that says: ‘what could I possibly being] delivered in the community, which aim to do — it’s too big/hard a problem for me, I have raise aspirations? little/no control over this — it’s always the same, this will just fail again’

Seeing (meaningful) change is absolutely crucial as there is a feeling that there have been no lasting changes in the past.

18 Suggested data sources

• Crime data

• Food bank data

• Data from credit scores, loans, rent, bills etc.

• Buying habits, supermarket purchase data

• The basic necessary income

• The link between finances and vulnerability

• School exclusion data

Potential responses

What interventions have previously been successful in raising aspirations of young people?

Can we model the key life decisions and what we can impact?

Potential implementation challenges and opportunities

• Needs sensitivity to the fact that many people are happy with their lives

• Anything focussed on schools (e.g. developing a curriculum that supports children’s aspirations) would need a Existing positive schemes huge amount of parental engagement — if parents are • Walk to work scheme — not involved in the process it positive results in the past is a wasted effort

• Lack of difficult to capture data — many people don’t interact with services

• Specific data with a high level of detail are needed

19 Project 5: Keep in mind there are different use natural experiments (differences mini-communities in Holme Wood detected between groups that do Adverse with different values, priorities etc. well and those that don’t) Childhood Link across generations (impact of Potential Experiences parents’ variables on childrens’), link ACEs and aspirations. implementation challenges Context — known and Lines of enquiry: the perceived issues economic case for action How can we identify ACEs from routinely collected data? This may require collecting a substantial ACEs (e.g. domestic violence, Can we demonstrate the resource amount of additional data. It is abuse, mental health of parents) and financial implications possible that schools could identify all have impact on the health and of providing/not providing many of the ACEs for children in outcomes for a child. intervention on other services — their schools but the collection of how do parents/carers and children such data are likely to be much Parents pass trauma on to children with ACEs interact with and tricker to obtain from parents/ indirectly as well as directly (e.g. navigate different services? carers and via self-report methods. genetics). Suggested data sources These problems don’t get solved — services are dealing with symptoms It would be useful to obtain not the underlying cause: rich longitudinal data to look at journey/pathway — link as many • Intergenerational gaps, existing data sets as possible, all transgenerational issues linked to a given person (need data (linked to other issues such as sharing agreements) education)

• Bad experience of parents May need qualitative data to has impact on their children supplement — why did these things (including biological and happen? physiological factors) Potential responses Lines of enquiry: • Link ACEs to health, defining the issue in education, transport, finances Holme Wood and crime • Model the link between How/when do you intervene? intergenerational ACE Need evidence for what works, experiences (i.e. how ACEs in need holistic view including health, parents affect lives of children). education, crime, housing, social • Create visualisations of ACE care, benefits, social networks/ data for specific communities influences (who is part of the children’s social network?), there Examine the evidence — which early should be data within public health targeted intervention can prevent and police etc. negative ACE outcomes (and are the best use of resources)? Can we

20 21 Annex C – Data scientists

22 The Digital Triangle: (individual, family, neighbourhood, etc.) to guarantee the robustness/completeness of the The three ‘apexes’ of the digital triangle comprise: modelling data analytics, artificial intelligence and immersive • A network of communication to address where technologies. Each reflects a transformative technology bottlenecks (e.g. economic or political) might exist in its own right, but the most revolutionary applications of data science for positive societal impact will The following areas were identified as real opportunities ultimately involve a combination of these technologies. to use data science techniques to test place based Thus, the triangle provides a powerful framework for working: Health Services; Education; Job Seeking; analyzing the challenges and opportunities that are Social Care; Transport. being created via this digital revolution. It was agreed across the data scientists that it will be The Alan Turing Institute is the National Centre for better understanding the upstream determinants of (e.g.) data analytics and artificial intelligence. Its support of health, and interactions between different services (e.g. this project brings internationally leading researchers in health and education) that will play to the strengths of the AI and data science to the partnership group. Research analytics capabilities offered by the Turing. For example, at the Turing Institute is channelled around a number understanding how physical and mental health impact on of ambitious challenges which represent areas in which an individual (and how social isolation can affect health AI and data science can have a game-changing impact and wellbeing) — exploring how nutrition affects dental for science, society, and the economy. The Home health and how dental health impacts on other aspects of a Wood project is seen as a great example of where data person’s life (e.g. attendance at school). science can be used to driving lasting societal benefit. It was agreed that it was the linking of various datasets The digital offering that would be the real game changer in the urban analytics approach offered by ActEarly: Holme Wood.

The data scientists identified the following skills and techniques in data science that could be brought to Data analytics bear on combined efforts to drive evidence based policy decisions — decisions with the potential to make lives The following points were made by the urban better for the people who live in Holme Wood: analytics experts:

• Machine learning techniques (Clustering, • There is a need to know what past research has happened Classification, Regressions) • A metadata catalogue needs to be created • Agent-based modelling • Time needs to be made available to clean the data • Other types of scientific modelling • Support is required to understand what the datasets • Visualization (to support viewing and interacting contain (for instance, what the data codes mean with data) for the variables of interest and how variables may interact with each other) • Causal inference (establishing what is causing what) • There is a need to link the data to get a full system The data scientists identified the following access to overview: Education, Health, Crime and Transport data or support was needed to enable data science to datasets would create a powerful platform for make a contribution to the shared goal: investigation

• Data from different local sectors (emphasis on • Context is key, and the data need to be interpreted transportation, schools, health and social services) in the light of ethnic and socio-economic backgrounds. This means that proper community • Linked databases representation is required to provide the ‘stories’ • Access to databases that evolve with time that allow for meaningful interpretation of the data

• Support for different types of data aggregation • It is the practitioners and community ‘on the ground’

23 who have the necessary • It would be good to create • It is often useful to use context‑specific knowledge simulation models: estimate visualisation to design models things in a different way (e.g. before using statistics to gain • There is always human bias in via agent-based modelling) outputs data analysis and this needs to be acknowledged (and appropriate • Machine learning would allow • These techniques can assist in controls put in place) the collation of qualitative the analysis of the data (e.g. results, and turn these into investigating data quality: • There is a need to identify from quantitative results to have a assessing whether there is the outset which stakeholders/ testable hypothesis sufficient data, seeing patterns clients are involved within any of missing data, showing given project • It will be important to make sure different sources, checking the solution fits the problem • Data scientists will need to compatibility of data sets) have facilitated contact with • There is a need to be aware • There is a need to allow a people in the right field to of the limitations, and project team to genuinely get the data that can answer a be transparent about the understand what data are specific research question constraints of the techniques available using visualisation • There is a need to intelligently • The visualisation process itself anonymise the data, thinking Data Visualisation can also stimulate an idea (e.g. about consent and ethical issues by observing unusual patterns) in advance The following points were made • It order to maximise the impact by the Visualisation experts: of this approach, there is a Artificial Intelligence need to know what attributes • Exiting visualisation of the data are important (e.g. techniques could help with The following points were made availability, time and space the interpretability and by the AI experts: granularity, names of fields, understanding of existing data/ example values coming from results (i.e. assist the way that • There is a need to bring fields, source of data). This information is communicated datasets together should reveal whether all to the community and policy needed data are available or • Temporal pattern mining makers) whether more data need to could be a particularly • These techniques can be collected. This can then effective approach be particularly useful in be presented in a way that’s • AI could provide descriptive communicating the unusual understandable analysis and help identify with regards to the average • There is a need to identify trends and statistics (that (something that may get lost in where this overview would sit could then be tested using reporting the statistics) (hosted on a website?) appropriate scientific methods) • The existing school admissions • It would be desirable if the • There is great scope to work is a good example of data visualisation tool could use predictive analysis: identifying unusual data be queried to a certain degree forecasting how a quantity patterns — with some schools (e.g. stratifying by age) changes in the future showing very abnormal patterns of intake • AI tools could allow for optimisation: if there are • These approaches can also limited resources it can be seen make the abstract personal — how to allocate them in the showing things in relatable best possible way ways that can inform policy making • There would be great promise in using causal interference

24 General comments

• It was noted that huge • It needs to be recognised that • The involvement of children amounts of data are constantly models might not provide and young people will allow generated, and substantial perfect solutions the development of role models infrastructure is required to on the Holme Wood estate host such datasets • There is need to use complementary techniques as • There is a need to make they tell us different things data science meaningful — and this means continually • It will be important to involve calibrating the findings with the community so they can the community understand the constraints of what is possible • It is important that projects are completed within the agreed • There is fantastic opportunity time frame to involve children and young people to show them how data • There is a good opportunity science is used to use the Wolfson Centre and take advantage of the existing expertise in digital science already deployed within the Born in Bradford project

• Studies should be designed in such a way to see that there can be confidence in determining whether trialled interventions have worked (or not)

• Where appropriate, scientists should be encouraged to develop methodologies that advance fundamental research on data analytics and AI and disseminate outputs to the scientific community (e.g. via academic publications) alongside impact-driven activities (e.g. building an evidence base for future government research evaluation exercises

• Geography and time are important factors that need to be captured

• There is an outstanding opportunity to train a ML/ AI model in one location (e.g. Holme Wood) then transfer it to learn another location

25

Looking for more information? Visit: www.caer.org.uk