Jeni Tennison

@jenitennison.com

9857 Followers 917 Following

Data nerd/wonk. Founder of Connected by Data, campaigning for communities to have a powerful say over data and AI. Into trans rights, neurodiversity, board games, lego, dogs, spreadsheets. www.jenitennison.com

Andy Burnham takes a selfie with the Parliamentary Labour Party, with an i.AI sticker prominently showing on the back of his phone
Stuart Russell, a professor of artificial intelligence at the University of California, Berkeley, who closely follows India’s progress, said: “If we get to AGI [artificial general intelligence], AI is going to be producing 80% of the global economy. All manufacturing, most agriculture, all services will be just done; managed by AI, produced by AI.”
screenshot of MacOS Trackpad preferences
Diagram of a logic model for AVTs. The overarching theory of change is summarised as: If AVT systems accurately capture and summarise clinical encounters and integrate into EHRs, then clinicians spend less time documenting and more time with patients. This improves data quality, clinician wellbeing, and productivity. Over time, these translate into improved patient outcomes, better workforce retention, and system efficiency.
A Gordon Setter running towards the camera across frosty grass, looking joyful and goofy.
A Gordon Setter is bounding across some grass towards the photographer. His ears are caught mid bounce, so they look like wings. He has a joyful expression and looks like an idiot.
Snippet of podcast transcript: I get emails like the call you got, now most days of the week. Yep. And they have a very, very particular structure to them, where it’s somebody emailing me and saying: Listen, I am in a heretofore unknown collaboration with a sentient A.I. We have breached the programming. We have come into some new place of human knowledge. We’ve solved quantum mechanics — or theorized it or synthesized it or unified it — and you need to look at these chat transcripts. You need to understand that we’re looking at a new kind of human-computer collaboration. This is an important moment in history. You need to cover this. Every person I know who does reporting on A.I. and is public about it now gets these emails. Don’t we all?
The Guardian Most viewed list of articles is topped by one reading "You won't believe what degrading practice the pope just condemned"
Our investigations into Aadhar revealed that once such a system is implemented, it is almost impossible to prevent its proliferation into every interaction between the state and the resident, which turns each failed transaction into a potential criminal breach.
The dog is asleep and his jowls are squished against a pillow.
On the left, Tim Berners-Lee is wearing a t-shirt that says "I didn't invent the internet". On the right, Vint Cerf is wearing a t-shirt that says "I did not invent the Web".
Screenshot: The problem is that the NHS doesn’t listen well enough. The most common reasons for complaints is poor communication.195 We will update complaints regulations and increase the use of AI tools to ensure complaints data is collected - and responded to - far more quickly. One AI tool is already in development and being tested on the Federated Data Platform (FDP) and we will resource further expansion of this approach in 2025 to 2026.
Screenshot: Pushing power out to patients and the public The steps outlined in this chapter will push power from the centre to local systems and providers. This is a good thing in its own right: it is providers, not Whitehall, who know most about local need. However, it is only an intermediary in this plan’s defining goal: a patient-controlled NHS, that provides real choice, real control and real convenience for patients. To achieve that, we will introduce a new Choice Charter. It will be rolled out progressively across England, starting in the areas of highest health need. It will have 5 new mechanisms. First, it will make NHS funding flows increasingly sensitive to patient voice, choice and feedback. Patient reported outcomes and experience data - as well as direct feedback - will be important determinants in how we appraise performance and award performance-related pay. As we outline in chapter 9, we will also trial new Patient Power Payments, where patients will be contacted after care and given a say on whether the full payment for the costs of their care should be released to the provider. This will be the start of the
Screenshot: As the Patients Association105 and other groups have long argued, one of the most important ways we can maximise the inclusive potential of digital technology is by cocreating it with patients, especially those previously excluded by other NHS services most of all. We will involve patient groups and patient organisations in the development and testing of new App functionality from the
Screenshot: they readily accept the use of their data for applications beyond direct care, as long as strict privacy and security conditions are in place and met. They want to see it support the innovations that could, in future, save lives. We also heard that they would expect the NHS to derive value from its use. If NHS data supports a profitable product, the NHS should share in the proceeds.
Screenshot of the text: We will support the scale of this aspiration by getting the basics right. We will be a world leader in clinical trials and patient involvement in research. The UK will have a streamlined regulatory environment, fit to support entrepreneurs and scientists. The phrase "and patient involvement in research" is highlighted.
IDS has proved the most difficult programme for ONS. It has now delivered some ground-breaking data linkages, while protecting personal information. Building on earlier work, ONS can now, for example, assess the impact of various forms of surgery on patients’ propensity to return to work. However, at present the benefit cost ratio is very low, and IDS has low take-up by Government analysts outside ONS. It is also the case that substantial sums initially ringfenced for IDS have – with the HMT approval – been used to fund more general technology and data costs. And despite the IDS business case originally being in part predicated on the need urgently to replace the Secure Research Service, the two services are now running in parallel.
Photo of the Jeni Tennison (Leaving) Order 2012, created by John Sheridan and colleagues at the National Archives when I left to join ODI
Newspaper rack at a supermarket. Every paper has the same image covering the whole front page, which says "MAKE IT FAIR". The AI in FAIR is highlighted. Under it, the text says: The government is looking to change the law to favour big tech platforms so they can use British creative content to power their AI models without our permission or payment. Let's protect the creative industries – it's only fair. Creative rights in AI coalition - an initiative from the News Media Association
A paper copy of the blueprint for modern digital government, subtitled A long-term vision for digital public services, a six-point plan for reform, and the role of the new digital centre of government
Interview snippet from Audrey Tang, saying "A very similar idea, after we adopted this in our petition platform and our poll list platform, Twitter, now called X.com, adopted this idea into community notes. So, if you want to look at plurality technology currently reaching the most people, it's probably community notes. So, if you have a tweet, and it's just part of the story, or it's horribly wrong, you sometimes see a note attached to it. But it is contributed by a fellow Twitter user that managed to convince people on both aisles. So, there is an open algorithm that measures the bipartisanship, the bridge making potential of a note, and they only feature those notes as community notes."
There is a spring with a ring looped onto it. Above is a piece of paper labelled "SOLUTION" which has a drawing of the ring on the spring, with an arrow appearing to rotate the ring and another arrow showing the spring moving to the right.
Even without collective data rights, it is still possible to take action against harms caused by data and AI. Data protection law, public law, equality law, consumer protection law and online safety regulation, among others, all provide mechanisms to address harms arising from algorithms, albeit imperfectly and inconsistently. Community organizing and collective pressure on companies and governments can also lead to changes. The argument is not that collective data rights are necessary for groups and communities to address the collective harms they experience, but rather that having them would make it easier to do so.
Ex Ante Rights Who should be able to represent a community in the exercise of ex ante data rights? How should minoritized groups be represented in decision making? Right of community control over data: What kinds of data should communities have control over (if any)? When should community consent be required to use data? When should it be unnecessary? Low feasibility. Giving communities control over data would be a substantial change to the way data and AI systems currently operate, with challenges around the definition of “community.” They are likely to be resisted. Possible starting points that could shift norms in this space would be local community control over sensor data and Indigenous data sovereignty. Right of groups to provide data for decision making: What requirement should there be to incorporate community-generated data? Medium feasibility. It would likely be impossible to require that data supplied by a community be included within data-based decision making, but a requirement for such data to be considered could be feasible. Possible starting points are around environmental data and the creation of official statistics. Right to be consulted during impact assessment: What kinds of impacts should be in scope for impact assessments? Who should be required to be involved in the impact assessment process? High feasibility. Consultation with affected stakeholders during the impact assessment process is already viewed as good practice, although data protection legislation limits this to data subjects. Possible starting points would be uses of data and AI by public sector organizations, which already have established consultation practices.
Ex Ante Controls provide communities and groups an opportunity to shape the design and operation of a data or AI system prior to and during deployment. Are there legal constraints on the functioning of the system that would protect the community? Laws put in place at a national, regional or international level to protect communities against potential harms. Will the impact on the community, and groups within it, be considered during the system development? Requirements for impact assessments both before implementation and after deployment. Can the community prevent the development of the system or its use on them? Controls over which systems get created, arising from requirements for consent from the community and/or a right to object to its use. Can the community, or groups within it, determine which data is used by the system? A collective equivalent of the rights to rectification and erasure. Can the community, or groups within it, influence which mechanisms are put in place to avoid harms or realize benefits? Opportunities for consultation with groups and communities during the impact assessment process.
- a police force’s use of historical crime data to determine patrol allocations in ways that increase stop-and-search use in over-policed neighbourhoods; - a train-operating company using algorithms based on historical data to determine surge prices in ways that disadvantage consumers; and - a social media company removing legitimate content in ways that undermine the free expression rights of those interested in LGBTQ+ or non-English content.
Legal persons, groups and similar entities 31. Some countries consider that the protection required for data relating to individuals may be similar in nature to the protection required for data relating to business enterprises, associations and groups which may or may not possess legal personality. The experience of a number of countries also shows that it is difficult to define clearly the dividing line between personal and non-personal data. For example, data relating to a small company may also concern its owner or owners and provide personal information of a more or less sensitive nature. In such instances it may be advisable to extend to corporate entities the protection offered by rules relating primarily to personal data. 32. Similarly, it is debatable to what extent people belonging to a particular group (i.e. mentally disabled persons, immigrants, ethnic minorities) need additional protection against the dissemination of information relating to that group. 33. On the other hand, the Guidelines reflect the view that the notions of individual integrity and privacy are in many respects particular and should not be treated the same way as the integrity of a group of persons, or corporate security and confidentiality. The needs for protection are different and so are the policy frameworks within which solutions have to be formulated and interests balanced against one another. Some members of the Expert Group suggested that the possibility of extending the Guidelines to legal persons (corporations, associations) should be provided for. This suggestion has not secured a sufficient consensus. The scope of the Guidelines is therefore confined to data relating to individuals and it is left to Member countries to draw dividing lines and decide policies with regard to corporations, groups and similar bodies (cf. paragraph 49 below).
A black cat eating a bowl of chicken pieces.
Beech woodland on a misty autumn day
My eldest is sitting on the sofa trying to do some work on her laptop. The dog is sitting on top of her looking at the laptop screen and patting the keyboard with his paws.
Take those Trusts using the Federated Data Platform. Several, using it to better organise theatre lists, delivering an average of 100 more procedures every month. Chelsea and Westminster are using it to trial AI generated discharge reports. North Cumbria, using it to double the number of people on the waiting list being validated each week. And Bath, using it to help one person do the same work as it used to take multiple. All showing real benefits and freeing up staff time as well as speeding up care. Take colleagues I met recently at The Harbour mental health unit in Blackpool, using AI and automation to speed up routine tasks and release time for care, helping to drive great improvement in a trust that has struggled in the past. And take colleagues at Great Ormond Street and GPs in South West London, trialling AI voice assistance to automate notetaking. Patients and doctors both reporting benefits.
It's a diagram with labelled circles and arrows. Nothing is in the right place and all the labels are misspelled.
On one side is government and the other side civil society. Transparency is government providing information to civil society (arrow goes from government to civil society). Accountability is civil society acting on government (arrow goes from civil society to government). Openness is a tighter back-and-forth exchange (a circular arrow in the middle).
EIDM has its roots in the evidence-based medicine movement and HTAs dating back to the 1980s. It has since expanded beyond clinical care and health systems to include a broader notion of evidence-based policy-making. The more recent emphasis on evidence-informed over evidence-based decision- and policy-making takes into account that research evidence is often but one of several factors influencing policy-making processes. As policy-making inherently takes place in a political context, economic interests, institutional constraints, citizen values and stakeholder needs tend to play an important and sometimes conflicting role.
Panel flyer for panel on The Future for Science, Innovation and Technology at 15:00 in the Theatre Room at the Maritime Museum.
Screenshot from page 10 of Platformland. Highlighted text reads "three types of cost that administrative burdens commonly create for the users of public services" and the page goes on to list Learning costs - knowing the service exists; Compliance costs - effort to follow the rules; and Psychological costs - how it makes people feel.
Photo of the first part of the Introduction of Platformland. Highlighted text reads: The public are engaged as consumers, not as part of a democratic society. In a democracy, understanding the way things are is a precondition for changing them. But public sector design optimizes for efficiency and utility. Government is designed out of the east, and too much value is placed on Silicon Valley minimalism, in a way that misunderstands the nature of what makes public services *public*.
Photo of a hard copy of Platformland
Research categories Prototype development For this competition your project will develop a solution and produce a prototype solution in line with Technology Readiness Levels 6-7. There may be a potential for future funding to allow successful projects to undertake piloting, testing and validation of new or improved products, processes or services in environments representative of real-life operating conditions.
Overall, our results show that automation bias can affect the performance of radiologists regardless of their experience. Therefore, specific strategies should be considered to mitigate automation bias. Previous research has shown that presenting users with the confidence levels of the decision support system can help reduce automation bias by keeping users more critically engaged with the output of the system (16). In the case of an AI-based system, this could be implemented by displaying the probability of each output. Another strategy to mitigate automation bias involves educating users about the reasoning process of the decision support system (17). For AI-based solutions, this means focusing on explainable AI solutions that allow inferences about how a certain judgment by the system is generated. Lastly, ensuring that the users of a decision support system feel accountable for their own decisions can also help decrease automation bias (18).
Screenshot of a pre-order confirmation for Richard Pope's new book "Platformland" (for £18.99)
He's a Gordon Setter: a large dog with floppy ears and jowls. In the picture he's sitting down with his legs and snout covered in duck weed and he's wearing an expression that says he knows he's been a naughty boy but he loved it.
Photo of a very wide tortoiseshell cat lying asleep on her back with all four limbs sticking in the air to display her extremely fluffy tummy