Let’s Bring Back Hope (p2): What Hope Means to the British Public
DDI United Kingdom · Audience response study
Part 2Let’s Bring Back Hope
What hope means, and what the public said it means. The reply to Prime Minister Andy Burnham’s “What Hope Means” video on Facebook, TikTok, Instagram and X.
Download the report (PDF)The full report, designed for print and reading offline
Part 2 of Let’s Bring Back Hope is one study published in three parts. All three draw on the same corpus of scored comments, and they are meant to be read together.
- ReportWhat Hope Means to the Public You are hereThe analytical report: findings, interpretation, and implications for policy and leadership.
- Interactive mapThe Discourse MapEvery scored comment on the map, with a guided reading, a comparison tool and seminar exercises.
- Lexical appendixThe Vocabulary of the ReplyThe words the public used, by platform, reception profile and DDI band.
Part 1: “Let’s Bring Back Hope” in 48 Hours, with its interactive discourse map and lexical appendix.
Key findings
- The ground has not moved. The mean DDI composite is 14.7 out of 100 and 79.7% of comments fall in the Risk band. Only 20 comments (0.16%) reach Healthy. Part 1 recorded 15.9 and 76.1% on a different instrument edition. The two readings are not a matched comparison, but they describe the same low ground.
- Restraint without generation. Commenters mostly held back from dehumanising and conspiratorial language: the four restraint indicators average 36.2% of their maximum. They rarely offered perspective-taking, solidarity, good faith or mobilisation: the sixteen generative indicators average 9.1%. More than one comment in four (28.1%) shows no generative trace at all.
- Five ways of answering hope. The indicator scores resolve into five reception profiles: Hostile (34.7%), Dismissive (26.4%), Civil scepticism (17.2%), Counter-demand (12.7%) and Constructive (9.1%).
- The public took the word back. 2,796 comments (23.0%) use the word hope. One in five of them denies that there is any; another fifth redefines it, most often as a general election, the removal of the government, control of the borders or relief on the cost of living. The most-replied-to comment in the whole corpus is a redefinition: “Hope is a General Election and a Reform Government”.
- The frame became the message. 349 comments mention the Prime Minister’s holiday, 78 the podcast microphone reading “How to Fail”, and 50 the music. Setting, soundtrack and timing were read as content.
- Platforms remain distinct grounds. Instagram (24.5) and TikTok (18.4) sit well above Facebook (12.8) and X (11.9). On X almost half the reply (47.7%) is Hostile; on Instagram more than a quarter (27.1%) is Constructive. The ordering holds when comments of similar length are compared.
- The best of the reply is particular. The highest-scoring comments are petitions and offers from people speaking out of a specific situation: a mental health crisis worker, a teacher, a retired counsellor, a parent explaining to a stranger how they got their son seen sooner.
ReHuman coded every included comment. A project-specific check against 19 human-reviewed comments found 79.19% exact agreement (293 of 370 indicator pairs). This is percentage agreement on a small reviewed sample, not Krippendorff’s alpha. All findings describe the captured corpus, not the British public.
Why hope, and the message
On 20 July 2026, in his first speech as Prime Minister, Andy Burnham asked the country to “bring back hope”. Seventeen days later he returned to the theme and tried to say what he meant.
Why this appeal matters to us
GCRD’s vision is a democracy in which citizens can engage with renewed hope. That is why Prime Minister Burnham’s appeal caught our attention, and why we have followed it through two studies. We find it particularly interesting and noteworthy because anti-democratic forces prefer a political atmosphere of hopelessness, and they cultivate it through a public discourse of hopelessness: the repeated message that every politician is the same, every reform is doomed and the future is closed. A leader who places hope at the centre of his first speech is, whatever one makes of his programme, contesting that ground.
As we argued in Part 1, hopelessness is a critical democratic vulnerability. It is the conviction that the future holds nothing that has not already been tried and failed. A citizen in that condition is not persuadable in the ordinary sense, because persuasion requires believing that evidence not yet seen could still change something. Demagogues speak fluently in the language of cynicism and despair, and in speaking it they deepen it. A politics of despair is never disappointed.
Europe has seen this before. In 1931 Marie Jahoda and Paul Lazarsfeld went to Marienthal, a factory village outside Vienna whose mill had closed, expecting to find the raw material of revolt. They found what they called a weary community (Jahoda, Lazarsfeld and Zeisel, 1933). Men walked more slowly; borrowing from the free library fell although they had nothing but time; political clubs emptied. Within two years Austria’s parliament had been suspended. Democracies do not fall only to the fury of crowds. They also fall to the fatigue of citizens who no longer believe that anything they do can matter.
Hope works the other way. It is a democratic force multiplier. Where despair subtracts, hope compounds: a community that organises successfully teaches others that organising works; a young voter who sees her ballot change something brings her friends next time; each act of agency, witnessed and shared, lowers the cost of the next. Hope in this sense is not optimism, which is a forecast the evening news can refute. It is the conviction that something is worth doing. Raymond Williams put the political consequence plainly: “It is then in making hope practical, rather than despair convincing, that we must resume and change and extend our campaigns” (Williams, 1989, p. 209).
Two spirals
Why the language of hope matters to democratic life. Hopelessness feeds on itself; so does hope.
A democratic vulnerability
Hopelessness
Back to step 1: each turn confirms the last.
A democratic force multiplier
Hope
Back to step 1: each turn recruits the next.
That is the question this study puts to the public reply. In what people wrote back to the Prime Minister, is hope being made possible, or is despair being made convincing?
From the appeal to the explanation
GCRD’s first report on the appeal, published in July, found a reception that was civil but closed. Across 8,086 coded comments the mean DDI composite was 15.9 out of 100, three comments in four sat in the Risk band, and the majority of responses fell in what social judgement theory calls the latitude of rejection (Sherif and Hovland, 1961). We argued then that the message was sound and the ground was poor: the problem lay less in what was said than in the condition of heart and the epistemic environment in which it was heard.
On 6 August the Prime Minister published a video across his four main channels. On X the caption read: “This is what bringing back hope means to me.” The clip was recorded for the How to Fail podcast, and the microphone in front of him bears the programme’s title (Figure 2). It was set to a music bed. It circulated during the parliamentary summer recess while, as many commenters pointed out, the Prime Minister was on holiday. A large part of the reply would not let that pass.
The frame of the message
A still from the “What Hope Means” video. The microphone cover reads “How to Fail”.
The move from Part 1 to Part 2 is analytically useful. Part 1 tested an appeal. Part 2 tests an explanation. When a leader meets scepticism, the instinct is to say more clearly what he means. This study asks whether explanation changes the condition of the ground on which a message lands.
The Democracy Discourse Index cannot measure persuasion, sentiment, voting intention or political support, and this report makes no claim about any of them. What it can describe is the democratic quality of the reply: whether people recognise the experience of others, disagree without contempt, extend trust to people, evidence and institutions, and speak as citizens who expect to act.
Method in brief
The DDI reads public discourse through four dimensions: empathy, civility, trust language and democratic agency. Each has five indicators, and every applicable indicator is scored on the same five-point anchor scale.
A comment’s composite is the sum of its active dimension scores expressed as a share of the points available, from 0 to 100. A dimension can be marked inapplicable; it then adds neither points nor possible points. Composites fall into four bands:
Band labels describe the score a comment receives. “Risk” is not a judgement that a commenter is dangerous, and a low score can mean the absence of a positive quality rather than the presence of hostility.
This study uses the Comment Edition of the DDI (Second Edition, September 2026), in which the unit of analysis is the comment. The Prime Minister’s post supplies context and is never scored.
Corpus
Comments were exported from the four source posts; the exports’ displayed dates run to 1 September 2026, from the day of posting on 6 August. Because the original X export held only 437 rows, we then retrieved the full reply conversation on X through a historical search. It returned 2,260 distinct replies before the results were exhausted, of which 1,835 were new to the corpus. Screening removed comments under 20 characters or three words, emoji-only comments, unresolved language and stimulus material.
From capture to analysis
Every stage keeps an explicit denominator.
Coding
Across the DDI programme, human triple-coding by locally grounded university partners provides the ground truth on which GCRD’s ReHuman AI system codes at scale and produces national scores. This audience response study applies ReHuman directly to a single-topic corpus. Because it rests on one message and a captured sample of the reply, its figures describe that reply and are not national DDI scores. Every included comment was coded, and the primary analysis uses the 12,139 comments coded in uniform ReHuman runs.
To check the coding within this study, ReHuman’s scores were compared with human-reviewed reference scores for 19 comments. They matched exactly on 293 of 370 indicator pairs (79.19%), came within one point on 93.78%, and differed by 0.286 points on average on the 0–4 scale. This is percentage agreement on a small sample that was neither random nor documented as blinded. It is not Krippendorff’s alpha and not an independent holdout test. Section 7 sets out what it does and does not support.
Further lenses used in this report
Four additional readings were built on the scored corpus for this report: reception profiles from k-means clustering of the twenty indicator scores; a transparent keyword dictionary of twelve themes; a rule-based classification of how comments use the word hope; and a thread analysis linking replies to the scores of the comments they answer. They are exploratory and descriptive, and none of them changes a DDI score.
Findings
3.1The same low ground
The mean composite across 12,139 comments is 14.7 and the median is 10. The distribution is heavily weighted toward the floor of the scale. 1,457 comments (12.0%) score exactly zero, and half score 10 or less. 79.7% sit in the Risk band, 17.2% in Concerning, 2.9% in Mixed and 0.16% in Healthy. Eight of the twenty Healthy comments were posted on Instagram, the smallest of the four audiences.
Most of the reply sits near the floor of the scale
Distribution of DDI composite scores, 12,139 comments. Bars are shaded by DDI band; darker means more positive democratic expression.
Show the data
| Band | Composite range | Comments | Share |
|---|---|---|---|
| Risk | 0–24 | 9,679 | 79.73% |
| Concerning | 25–49 | 2,093 | 17.24% |
| Mixed | 50–74 | 347 | 2.86% |
| Healthy | 75–100 | 20 | 0.16% |
Part 1 measured the first appeal with an earlier edition of the instrument, a different coding model and a 48-hour window. Part 2 measures a different message over nearly four weeks with the Comment Edition and ReHuman. We therefore do not read the difference between 15.9 and 14.7 as change. We read the similarity as information. Two messages, two instruments and two collection designs return the same picture: roughly four comments in five in the Risk band, X at the bottom, Instagram and TikTok at the top.
One movement inside Part 2 should not be misread either. Adding the full X conversation lowered the pooled mean from 15.17 to 14.69, because the 1,580 added replies average 11.45. That is a change in the composition of the corpus, not evidence that the discourse deteriorated over time.
Two messages, two measurements, the same ground
Mean DDI composite in Part 1 (the “Let’s bring back hope” appeal, July 2026) and Part 2 (the “What Hope Means” video, August 2026). This is a side-by-side reading, not a matched comparison.
Show the data
| Platform | Part 1 mean | Part 2 mean | Part 1 Risk | Part 2 Risk |
|---|---|---|---|---|
| All platforms | 15.9 | 14.7 | 76.1% | 79.7% |
| 15.0 | 12.8 | 79.1% | 84.0% | |
| TikTok | 20.3 | 18.4 | 63.3% | 72.3% |
| 20.1 | 24.5 | 64.0% | 55.9% | |
| X | 13.2 | 11.9 | 83.2% | 85.4% |
3.2Restraint without generation
Civility is the strongest dimension at 19.7% of its maximum, followed by trust language (14.2%), empathy (12.5%) and democratic agency (11.8%). Even the strongest dimension reaches barely a fifth of what the scale allows.
Civility leads, agency trails
Mean dimension score as a share of its maximum (20 points). The figure on the right is the mean out of 20.
The dimension averages hide a sharper pattern. Four of the twenty indicators register the absence of harm: anti-hostility, respectful tone, non-dehumanisation and anti-conspiracy. We call these restraint indicators. The other sixteen register a positive contribution: taking another’s perspective, recognising emotion, standing with others, disagreeing constructively, de-escalating, reading others generously, extending good faith, trusting institutions and evidence, and expressing efficacy, mobilisation, rights or democratic optimism. We call these generative indicators.
Read this way, the reply has a clear shape. The restraint indicators average 36.2% of their maximum; the generative indicators average 9.1%. Non-dehumanisation (1.88 of 4) and anti-conspiracy (1.58) are the only indicators whose median reaches the Moderate anchor. The least expressed are emotional recognition (0.09), perspective-taking (0.11), generous interpretation (0.15), good faith attribution (0.21) and mobilisation (0.24).
Restraint is common; generation is rare
Mean score on each of the twenty indicators (0 = absent, 4 = exemplary). Axis runs from 0 to 2.
Show the data
| Indicator | Type | Mean (0–4) | % of maximum |
|---|---|---|---|
| Perspective-Taking | Generative | 0.11 | 2.7% |
| Emotional Recognition | Generative | 0.09 | 2.3% |
| Humanisation | Generative | 0.66 | 16.4% |
| Solidarity | Generative | 0.47 | 11.8% |
| Anti-Hostility | Restraint | 1.18 | 29.4% |
| Respectful Tone | Restraint | 1.18 | 29.4% |
| Non-Dehumanisation | Restraint | 1.88 | 46.9% |
| Constructive Disagreement | Generative | 0.39 | 9.7% |
| De-escalation | Generative | 0.34 | 8.6% |
| Generous Interpretation | Generative | 0.15 | 3.8% |
| Institutional Trust | Generative | 0.49 | 12.1% |
| Good Faith Attribution | Generative | 0.21 | 5.2% |
| Interpersonal Trust | Generative | 0.29 | 7.1% |
| Epistemic Trust | Generative | 0.29 | 7.2% |
| Anti-Conspiracy | Restraint | 1.58 | 39.4% |
| Political Efficacy | Generative | 0.82 | 20.6% |
| Mobilisation | Generative | 0.24 | 6.1% |
| Framing Collective Action | Generative | 0.48 | 12.0% |
| Rights Consciousness | Generative | 0.35 | 8.8% |
| Democratic Optimism | Generative | 0.47 | 11.6% |
The same pattern appears at the level of the single comment. 12.0% score zero on every indicator. A further 16.1% show restraint and nothing else. Only 13.7% reach the Strong anchor on any generative indicator.
How far does a comment reach beyond restraint?
Each comment is placed by the highest score it reaches on any of the sixteen generative indicators. Comments with all four dimensions active.
Show the data
| Platform | Nothing at all | Restraint only | Weak generative trace | Moderate generative trace | Strong generative expression |
|---|---|---|---|---|---|
| All platforms | 12.0% | 16.1% | 26.0% | 32.2% | 13.7% |
| 14.0% | 17.4% | 26.9% | 30.6% | 11.0% | |
| TikTok | 4.9% | 15.1% | 26.3% | 36.1% | 17.6% |
| 7.6% | 7.5% | 17.1% | 35.8% | 32.0% | |
| X | 18.5% | 16.1% | 24.8% | 29.8% | 10.9% |
Most of this public, in other words, is not abusive. Citizens were largely willing not to dehumanise the Prime Minister. They were rarely willing to imagine his position, or one another’s. Part 1 described the reception as civil but closed. Part 2 gives that phrase a measurable form: a civil floor with very little built on it.
3.3Five ways of answering hope
Averages describe the middle of a distribution that has very little middle. To see the different ways people actually replied, we clustered the twenty indicator scores of the 12,111 comments on which all four dimensions were active. Five reception profiles emerge. They are tendencies with soft edges rather than discrete types, and they describe comments, not commenters.
Five ways of answering hope
Reception profiles from k-means clustering of the twenty indicator scores. Each circle is a profile, placed by its average relational quality and democratic agency; area shows its share of comments.
Show the data
| Profile | Comments | Share | Mean composite | Empathy | Civility | Trust | Agency |
|---|---|---|---|---|---|---|---|
| Hostile | 4,198 | 34.7% | 2.7 | 1.1% | 2.0% | 3.0% | 4.6% |
| Dismissive | 3,192 | 26.4% | 9.9 | 7.5% | 18.4% | 10.3% | 3.0% |
| Counter-demand | 1,537 | 12.7% | 20.0 | 9.9% | 20.9% | 17.1% | 31.3% |
| Civil scepticism | 2,084 | 17.2% | 24.8 | 27.4% | 37.1% | 23.3% | 10.9% |
| Constructive | 1,100 | 9.1% | 47.2 | 45.9% | 55.9% | 47.1% | 39.6% |
Hostile
34.7%of comments · mean composite 2.7
Empathy 1% · Civility 2%
Trust 3% · Agency 5%
Near-zero on every indicator, including the restraint floor. This is contempt, abuse and scapegoating, often directed at the Prime Minister personally. The integrity screen flags about a third of these comments (33 to 35 per cent) for suspected misinformation, mostly unsubstantiated allegation.
Abstract concepts. f*** off.
A FRAUDULENT PRIME MINISTER CAN’T BELIEVE WHAT BRITISH POLITICS HAS BECOME
Dismissive
26.4%of comments · mean composite 9.9
Empathy 8% · Civility 18%
Trust 10% · Agency 3%
Refusal without abuse. These comments do not dehumanise, but they offer nothing beyond the refusal: “word salad”, “all talk”, “more of the same”. Two in five of the comments about the Prime Minister’s holiday sit in this profile.
Yet another Labour word salad which is meaningless.
All he does is talk talk talk, Actions speak louder than words. This countries down the Toilet.
Counter-demand
12.7%of comments · mean composite 20.0
Empathy 10% · Civility 21%
Trust 17% · Agency 31%
Agency without recognition. Political efficacy and collective-action framing are present, empathy is almost absent. 48% of these comments mention an election or a vote. Hostile to the messenger, they still make a claim on the democratic process.
CALL A GE & RESIGN, that'll give me hope.
We have seriously lost hope with your government. call an election that might give us hope
Civil scepticism
17.2%of comments · mean composite 24.8
Empathy 27% · Civility 37%
Trust 23% · Agency 11%
Courteous, measured doubt. Civility and some empathy are present, agency is low. These commenters are not persuaded and say so without contempt: hope must be earned, delivered, shown.
We can have hope, doesn’t mean Labour can supply it, haven’t done a good job so far
Hope needs to be earned.
Constructive
9.1%of comments · mean composite 47.2
Empathy 46% · Civility 56%
Trust 47% · Agency 40%
The full democratic register. Every dimension is expressed, and comments are about twice the median length of the others. They petition, propose and offer, usually from a particular circumstance.
I am clinging on by my finger nails that you listen to the teaching profession and really address our concerns: Ofsted, funding, new curriculum, reduce the workload. PLEASE!🙏
You want to bring back hope ? Give careful thought to th benefits process. Some of us really are stuck between a rock and a hard place
Where each profile lives
Share of each platform’s comments in each reception profile.
Show the data
| Platform or type | Hostile | Dismissive | Counter-demand | Civil scepticism | Constructive |
|---|---|---|---|---|---|
| 39.2% | 26.3% | 12.9% | 15.6% | 6.1% | |
| TikTok | 19.8% | 29.5% | 14.6% | 23.3% | 12.7% |
| 24.1% | 17.6% | 11.7% | 19.5% | 27.1% | |
| X | 47.7% | 23.9% | 9.2% | 11.8% | 7.4% |
| All replies | 27.7% | 29.1% | 6.6% | 25.2% | 11.4% |
| All top-level | 36.2% | 25.8% | 14.0% | 15.5% | 8.6% |
Two features of the profile distribution matter for what follows. First, X and Facebook are dominated by the Hostile profile (47.7% and 39.2%), while on Instagram the Constructive profile is the largest single group (27.1%). Second, when commenters answer one another rather than the Prime Minister, the mix changes. Civil scepticism rises from 15.5% of top-level comments to 25.2% of replies, Hostile falls from 36.2% to 27.7%, and Counter-demand more than halves, from 14.0% to 6.6%. In conversation with each other, citizens become more civil and less demanding.
Civil scepticism, Counter-demand and Constructive together make up 39.0% of the reply. Part 1 identified a conditional segment, commenters who “acknowledge the appeal while refusing unconditional belief”. Part 2 did not code stance, so the two cannot be matched, but the Civil scepticism and Counter-demand profiles describe the same kind of public: unconvinced, still talking, and specific about what would change their minds.
Every comment behind these five profiles can be explored in place in The Discourse Map, the interactive companion to this report.
3.4What the public says hope means
The Prime Minister’s video was an answer to the question of what hope means. The reply answered the same question. 2,796 comments (23.0%) use the word hope, rising to 31.5% on X. These comments score higher on average than those that do not (17.1 against 14.0): taking up the leader’s word, even to contest it, is a form of engagement.
What commenters did with the word
How the 2,796 comments that use the word hope use it. Bars show share; the right-hand figure is the mean DDI composite of each use.
Show the data
| Use of hope | Comments | Share | Mean composite | Risk band |
|---|---|---|---|---|
| Redefined: “hope is…” | 615 | 22.0% | 18.6 | 70.2% |
| Denied: “there is no hope” | 571 | 20.4% | 11.3 | 88.3% |
| Wished: “I hope…”, “let’s hope” | 393 | 14.1% | 19.7 | 69.2% |
| Quoted or mocked | 102 | 3.6% | 15.2 | 79.4% |
| Other use | 1,115 | 39.9% | 18.5 | 71.7% |
How the word is used matters more than whether it appears. Comments that deny there is any hope score lowest (11.3, with 88.3% in Risk). Comments that redefine it score 18.6, and wishes score 19.7. To argue about what hope means is already to take part in a public conversation. To declare that hope does not exist is to leave it.
The only hope you're bringing back is no hope.
You sir are deluded.. hope is a feeling, and feeling isn’t a solution..
Hope isnt enough,we need much more than that.
Bring back hope by calling for a General Election!
What the public said hope would take
Conditions mentioned in the 615 comments that redefine hope. A comment can mention more than one.
Show the data
| Condition mentioned | Comments | Share of redefinitions | Mean composite |
|---|---|---|---|
| A general election | 162 | 26.3% | 22.8 |
| Removal of the government or PM | 89 | 14.5% | 12.3 |
| Borders, boats and deportation | 84 | 13.7% | 12.1 |
| Cost of living and tax | 77 | 12.5% | 20.6 |
| Reform or Restore in power | 61 | 9.9% | 24.0 |
| Delivery, not words | 55 | 8.9% | 25.8 |
| Public services and welfare | 44 | 7.2% | 23.4 |
| Gaza and foreign policy | 22 | 3.6% | 16.9 |
The redefinitions read like terms of settlement. The single most common condition is a general election (26.3% of redefinitions), followed by the removal of the government or Prime Minister (14.5%), the borders and deportation (13.7%), the cost of living (12.5%), a Reform or Restore government (9.9%), delivery rather than words (8.9%) and public services (7.2%).
Here the DDI registers something easy to miss. A demand for an election is hostile to the messenger and faithful to the mechanism. It scores as democratic agency, and rightly: it is a claim on the democratic process itself. Many of these comments tie the demand to the way the Prime Minister came to office, with phrases such as “no mandate”, “unelected” and “not voted in”. Redefinitions that call for an election average 22.8, above the corpus mean. Those that point to removal (12.3) or the borders (12.1) sit well below it. Part of this public is defining hope as consent.
The ‘Hope’ is in the shape of a General Election! 🗳️👎Blah Blah Blah..
Hope would be a general election where you can change the PM, but at the end of the day, the Labour Party just can't be trusted to talk a lot, but as usual, nothing is done about the true problems […]
STOP THE BOATS AND START DEPORTATIONS OTHERWISE ALL YOUR HOPE WILL BE EXTINGUISHED ALONG WITH YOUR PARTY
Hope looks like:-Government backed mortgages at a fixed 2.5%.Retroactive student loan interest rates of a fixed 0.5%.Subsidised Fruit and Veg in shops of 75%.Building affordable homes not council […]
The most-replied-to comment in the corpus was a redefinition posted on Facebook: “Hope is a General Election and a Reform Government 👌”. It drew 377 replies, more than any other comment. Those replies average 12.3, and 27% of them mention Reform or Farage, for and against. At this point the argument about hope is no longer between the Prime Minister and his public. It is between publics.
3.5What the reply talks about
Themes carry quality. A transparent keyword dictionary finds that the most frequent subjects are the political class (14.2% of comments), the cost of living and tax (13.0%), immigration and borders (12.2%), elections and the mandate (11.3%) and the gap between words and action (10.5%).
Subjects carry quality
Twelve themes from a transparent keyword dictionary, ordered by how often they appear. Dots show the mean DDI composite of comments on each theme; the figure on the right gives the share of all comments.
Show the data
| Theme | Comments | Share | Mean composite | Risk band | TikTok | X | ||
|---|---|---|---|---|---|---|---|---|
| Labour, Starmer and the political class | 1,727 | 14.2% | 12.3 | 85.9% | 15.2% | 12.8% | 10.6% | 14.5% |
| Cost of living and tax | 1,583 | 13.0% | 18.9 | 71.3% | 11.5% | 16.9% | 14.9% | 11.1% |
| Immigration and borders | 1,478 | 12.2% | 10.6 | 90.7% | 14.7% | 6.7% | 7.1% | 14.3% |
| Election and mandate | 1,371 | 11.3% | 19.0 | 74.3% | 12.8% | 10.6% | 6.2% | 9.1% |
| Words versus action | 1,276 | 10.5% | 14.3 | 81.3% | 10.2% | 10.1% | 11.3% | 11.7% |
| Public services and welfare | 879 | 7.2% | 21.4 | 65.4% | 7.7% | 6.7% | 12.2% | 5.2% |
| Reform and the insurgent right | 634 | 5.2% | 18.1 | 73.2% | 6.8% | 4.6% | 3.0% | 1.9% |
| Gaza, Israel and foreign policy | 423 | 3.5% | 13.7 | 80.4% | 2.8% | 2.6% | 17.2% | 2.9% |
| Free speech and policing | 266 | 2.2% | 16.4 | 77.1% | 1.9% | 1.7% | 5.4% | 2.9% |
| Climate, energy and technology | 241 | 2.0% | 24.8 | 56.4% | 1.2% | 2.6% | 9.8% | 1.1% |
| Manchester and the mayoral record | 221 | 1.8% | 14.5 | 81.0% | 2.1% | 1.6% | 0.8% | 1.6% |
| Well-wishing and support | 153 | 1.3% | 30.3 | 44.4% | 1.1% | 1.4% | 3.9% | 0.8% |
Comments on immigration and borders score lowest (10.6, with 90.7% in Risk), followed by the political class (12.3) and Gaza, Israel and foreign policy (13.7). Comments on climate, energy and technology (24.8), public services and welfare (21.4), elections (19.0) and the cost of living (18.9) score above the corpus mean. Subjects grounded in lived circumstance, such as a bill, a hospital or a child’s school, carry more of the generative register than subjects framed as threat or betrayal. The lexical appendix, The Vocabulary of the Reply, sets out the words behind these themes by platform, reception profile and DDI band.
The frame as message
A substantial share of the reply was about the video rather than its argument. 349 comments mention the Prime Minister’s holiday or the summer recess (mean 12.6). 78 notice that the microphone in front of him bore the podcast’s title, “How to Fail” (Figure 2; mean 11.9). 50 object to the music (mean 9.9). McLuhan’s old observation that the medium is the message (McLuhan, 1964) holds on new ground. On low-trust terrain, the setting, the soundtrack and the timing are part of what is said, and they are often heard first.
Why does the microphone say "How to Fail"? I'm in favour of the optimism Mr B brings to the table but surely this is a video faux-pas
I'm tired of this posturing. You're off on holidays while the rest of us counting pennies. Cmon. Stop pretending to be a man of the people.
Please drop the sad music, drop the act. I appreciate this man is trying to correct the failures of kier starmer. But hes not going to be the man to […]
Hope is not having podcasts we used to enjoy taken over by political parties for their own PR purposes.
Defending the holiday came at a price. A Facebook comment asking readers to “give him a fair chance”, which scored 51, attracted 50 replies averaging 15.4.
3.6Four platforms, four grounds
Instagram returns the highest mean (24.5, with 55.9% in Risk), then TikTok (18.4, 72.3%), Facebook (12.8, 84.0%) and X (11.9, 85.4%). The ordering runs through every one of the four dimensions.
Four platforms, four grounds
Mean composite, Risk-band share, dimension scores (% of maximum) and the two outer profiles, by platform. Darker cells indicate more positive democratic expression.
| Platform | Comments | Mean composite | Risk band | Empathy | Civility | Trust | Agency | Hostile profile | Constructive profile |
|---|---|---|---|---|---|---|---|---|---|
| 6,361 | 12.8 | 84.0% | 10.4% | 17.2% | 12.4% | 10.7% | 39.2% | 6.1% | |
| TikTok | 3,224 | 18.4 | 72.3% | 16.4% | 25.4% | 17.1% | 14.0% | 19.8% | 12.7% |
| 592 | 24.5 | 55.9% | 23.4% | 30.3% | 24.1% | 19.6% | 24.1% | 27.1% | |
| X | 1,962 | 11.9 | 85.4% | 9.7% | 15.3% | 12.3% | 9.8% | 47.7% | 7.4% |
Swipe the table sideways to see every column.
A natural objection is that Instagram and TikTok comments are simply longer and so have more room to express positive qualities. Length does matter: the composite rises steadily with word count. But the platform ordering survives inside every length band. A 21–40 word comment on Instagram averages 26.2; one of the same length on X averages 13.5.
The platform ordering survives a length check
Mean DDI composite by comment length and platform. Longer comments score higher everywhere, but Instagram and TikTok stay above Facebook and X within every length band.
| Words in comment | TikTok | X | All | ||
|---|---|---|---|---|---|
| 3–5 | 9.7 | 13.7 | 15.3 | 7.6 | 10.6 |
| 6–10 | 11.2 | 15.2 | 16.6 | 11.3 | 12.5 |
| 11–20 | 12.1 | 18.4 | 21.8 | 11.9 | 14.1 |
| 21–40 | 13.9 | 22.4 | 26.2 | 13.5 | 16.8 |
| 41–80 | 17.5 | 25.7 | 33.4 | 15.0 | 19.9 |
| 81+ | 23.3 | 27.1 | 34.7 | n < 20 | 25.7 |
The platforms also talk about different things. Instagram’s reply is organised around Gaza and Israel (17.2%), public services (12.2%) and climate, energy and technology (9.8%). Facebook and X carry the immigration argument (14.7% and 14.3%) and the argument about the political class. TikTok leans toward the cost of living and tax (16.9%).
X again records the lowest composite of the four platforms, as it has across GCRD’s monitoring. We read this as the primary implication of the platform finding rather than an artefact of sampling. X is where political journalists, opinion leaders and engaged partisans concentrate, and where older users are well represented; Facebook skews older and toward community groups; younger audiences on TikTok and Instagram appear more open and more hopeful. This study collected no demographic data, so these are readings of the platforms’ known publics rather than measured attributes of these commenters. X is also where hope is most often quoted (31.5% of comments) and where the integrity screen flags the most suspected misinformation (19.8% of included X comments).
3.7The reception in motion on X
This section uses the supplementary X retrieval, the one collection with API timestamps when the analysis was frozen, so it is about X alone. The median reply arrived 5.8 hours after the post. Half had arrived within six hours and three quarters within a day. Through the first 48 hours the mean composite stayed between 9.9 and 11.4.
On X, the verdict is fast and it holds
Replies to the X post by time of arrival, and their mean DDI composite. Supplementary full-conversation retrieval only (n = 1,580), the one collection with API timestamps when this analysis was frozen.
Show the data
| Window | Replies | Mean composite | Risk band |
|---|---|---|---|
| 0–3h | 531 | 10.8 | 87.4% |
| 3–6h | 271 | 9.9 | 89.7% |
| 6–12h | 152 | 10.5 | 86.2% |
| 12–24h | 251 | 11.4 | 87.3% |
| 24–48h | 139 | 11.1 | 85.6% |
| 2–4 days | 99 | 18.9 | 70.7% |
| 4–7 days | 100 | 13.6 | 84.0% |
| 7+ days | 37 | 11.5 | 86.5% |
Part 1 found reception hardening as volume peaked and softening in the tail. On X in Part 2 there is no softening in the window when most people reply. The verdict is fast and it holds. A brief rise at two to four days rests on 99 replies and fades by the end of the week.
Collection method shapes what a platform appears to say. The 382 X replies in the original export average 13.6; the 1,580 added by full-conversation retrieval average 11.45. Exporters may favour more visible replies. The full conversation is harsher than its most visible part.
3.8When citizens answer each other
Replies to other commenters score higher than comments addressed to the post (17.3 against 14.1). The pattern holds on Facebook, TikTok and X; Instagram, where top-level comments are already the strongest, is the exception.
When citizens answer each other
Mean DDI composite of top-level comments (answering the post) and replies (answering another commenter), by platform.
Show the data
| Platform | Top-level n | Top-level mean | Reply n | Reply mean |
|---|---|---|---|---|
| All | 9,952 | 14.1 | 2,187 | 17.3 |
| 5,093 | 12.0 | 1,268 | 15.9 | |
| TikTok | 2,624 | 17.9 | 600 | 20.4 |
| 454 | 25.1 | 138 | 22.6 | |
| X | 1,781 | 11.8 | 181 | 12.9 |
Tone travels down a thread, though weakly. Replies to comments scoring 50 or more average 24.1; replies to comments scoring 0–4 average 13.6.
Better comments draw somewhat better replies
Mean DDI composite of replies, grouped by the score of the comment they answer (1,949 replies matched to a scored parent).
Generosity is not self-protecting. On X, a reply praising the Prime Minister for “decency, purpose and a genuine belief that Britain can be better than this” scored 48 and drew 47 replies averaging 7.9, with 89% in Risk. Endorsement on this ground attracts attack. Yet the same corpus holds the clearest picture of what the DDI is looking for: a stranger answering another parent’s worry about a long wait for a child’s diagnosis.
This is exactly why Andy Burnham gives people hope. No empty theatrics, no manufactured outrage, just decency, purpose and a genuine belief that Britain can be better than this.He understands that leadership is about giving people something to believe in again. 🌹
I don’t know if this will help but I was in the same situation with my son. I contacted my local mp and he was very helpful in getting my son seen earlier. Some good news for the future is that GP’s are training to be able to do this diagnosis locally. Hopefully this will reduce the awful wait times. Best of luck to you. Xx
3.9The narrative grammar of grievance
Alongside the DDI, the coding pipeline runs a separate information-integrity screen. It labels 1,846 of the 12,172 included comments (15.2%) as suspected misinformation, disinformation or malinformation: 74% for unsubstantiated allegation and 25% for misleading framing. The labels are screen-level suspicions. Calling a politician a “puppet” registers as an allegation; no claim was fact-checked, and no prevalence of actual misinformation is estimated. What the screen does show is the grammar of the grievance.
The narrative grammar of grievance
Meta-narratives the integrity screen detects in the 1,846 included comments carrying a suspected misinformation, disinformation or malinformation label. A comment can carry several.
Two thirds of flagged comments rest on a story of corrupt elites, and nearly half on blaming an out-group. Economic decline and threatened national identity follow. Flagged comments average 5 to 6 on the DDI composite, against about 16 for the rest. This is the pessimistic age in miniature: the people at the top are corrupt, the people at the edges are to blame, and the nation is in decline. It is a grammar in which hope can only arrive as removal.
Interpretation: what explanation cannot do
A leader can clarify a word. He cannot, by clarification alone, change the condition in which the word is heard.
Where hope goes
A reading of the study as a sequence, from the message to the openings it leaves.
“This is what bringing back hope means to me”
A video on four channels, 6 August 2026, seventeen days after the first appeal.
Read before the words
- Holiday: 349 comments
- “How to Fail” microphone: 78
- Music: 50
Restraint without generation
- Restraint 36% of maximum, generation 9%
- Flagged comments: corrupt elites 67%, scapegoating 48%
Hope taken back
- Hostile 35%, Dismissive 26%
- Civil scepticism 17%, Counter-demand 13%
- Constructive 9%
Where leadership can still meet the public
- Answer the petitioner
- Make hope checkable
- Respond to the demand for consent
Explanation does not reach the ground
The second message was framed in the first person, as what hope means “to me”, and was meant to say what the first had left open. The reply looks much as it did in July. That does not show the explanation failed on its own terms; the DDI does not measure persuasion. It shows that the democratic quality of the reply is a property of the ground more than of the message. Humans do not settle for truth; they settle where they belong. Part 2 suggests that they also reply from where they belong. The Hostile and Dismissive profiles are not reactions to this video so much as positions held in advance.
Hope is a contested property
When a leader invokes hope, he does not own the word. The public took it back and filled it with conditions: an election, the borders, the bills, the removal of the government. This is not the absence of hope. It is hope stated as a contract, with terms and a test. “Hope needs to be earned” is a sceptical sentence, but it is also a hopeful one: it assumes that earning is possible.
Restraint is not recognition
The four-to-one ratio between restraint and generation is the most important number in this report. It describes a public that has largely learned not to dehumanise, but has not been invited into, or has withdrawn from, the harder work of recognising others. Democracies can absorb anger. They struggle with the absence of recognition, because recognition is what makes disagreement bearable and compromise imaginable.
The democratic form of the demand
The Counter-demand profile is the most politically instructive. It is hostile to the messenger and loyal to the mechanism. A public that answers “what hope means” with “a general election” is making a legitimacy claim about a premiership that began without one. Treating this as noise would be a mistake; it is one of the few places where the reply expresses democratic agency at all.
The particular is where democracy breathes
The Healthy and Constructive comments share a grammar. They speak from an identifiable situation: a crisis team, a classroom, a benefits claim, a son on a waiting list. They address the Prime Minister as someone who could act, and each other as people worth helping. Rehumanisation begins where a citizen is heard as a person with a case. This is also where the Five-Tier Trust Architecture locates the restoration of trust: in relationship before institution, and in visible faithfulness before rhetoric.
Implications for policy and leadership
These implications follow from the descriptive findings. They are offered to government, to political communicators and to the platforms, and they do not depend on any judgement about the Prime Minister’s programme.
Answer the petitioners
The Constructive profile is 9.1% of the reply overall and 27.1% on Instagram. These comments bring specific cases and proposals. A visible protocol for acknowledging, routing and reporting back on them would do more for trust than another message about hope. The public asked for exactly this.
Want people to feel heard? Start with your constituency emails.
X · CommentDDI 46Constructive Another good post, Prime Minister. But before you post again, please respond to the replies and emails from me and the many others who have responded to your post and also emailed you directly. Prime Minister, you’ve got mail! 😃
Instagram · CommentDDI 54Constructive Make hope checkable
The public defined hope as delivery: an election, the borders, the bills, the services. Hope that can be audited is the only kind this ground will accept. A small number of testable commitments, reported against in public at regular intervals, would turn a contested word into a shared measure.
Respond to the demand for consent
The most common counter-definition of hope was a general election. Engaging the question of mandate directly, and explaining the constitutional basis of the premiership with respect for those who doubt it, treats the counter-demand as the democratic claim it is.
Treat the frame as part of the message
Timing, music and setting were read as content by hundreds of commenters. On low-trust ground, production values read as performance. Messages about hope should be timed and staged with the same care as their words.
Speak to each platform as its own public
X is the hardest ground and the platform where opinion leaders concentrate; it needs argument and evidence. Facebook carries half the volume and the community arguments about immigration and the political class. Instagram and TikTok hold the most constructive replies and the cost-of-living and public-service agenda.
Protect the generous
Supportive and fair-minded comments draw pile-ons. Platform design, community moderation and the example of public figures who engage with constructive replies can make generosity less costly.
Listen at scale, and validate independently
The restraint-to-generation ratio, the profile mix and the narrative grammar of grievance are measurable over time. Used alongside polling, the DDI can track whether public discourse recovers recognition and agency, and flag patterns that warrant closer investigation. That use requires larger, independent and blinded human validation than this audience response study carries.
Hope that can be checked
The Prime Minister asked what hope means, and the public answered. The answer was partial, frequently angry, sometimes cruel, and often procedural. Underneath it lies something that deserves attention. Large numbers of people still believe things could be different if certain conditions were met. Their conditions are not ours to approve. But a public that sets out conditions has not given up on democratic life.
The data trace a movement worth noticing. Denial of hope scores lowest. Redefinition scores higher. Petition scores highest. The path from “there is no hope” to “hope would be…” to “here is my case, please look at it” is the path democratic leadership has to make possible. It cannot be walked by communication alone. It needs listening that leaves a trace, restraint that protects the generous, and faithfulness to promises that people can check for themselves.
The Rule of Saint Benedict opens with a single instruction: listen, and listen “with the ear of the heart” (Rule of Saint Benedict, Prologue 1). Part 1 of this study ended with the condition of the heart in which hope is received. Part 2 ends with the ear. The public has said, in thousands of imperfect sentences, what hope would mean to them. Hope comes back where people are heard, and where what they are promised can be seen.
Hope comes back where people are heard, and where what they are promised can be seen.
Methods, limits and data
Sources and capture
| Platform | Captured | Included | Primary n | Mean | Risk band |
|---|---|---|---|---|---|
| 8,012 | 6,378 | 6,361 | 12.8 | 84.0% | |
| TikTok | 4,162 | 3,234 | 3,224 | 18.4 | 72.3% |
| 703 | 592 | 592 | 24.5 | 55.9% | |
| X | 2,272 | 1,968 | 1,962 | 11.9 | 85.4% |
The original sources were four ExportComments workbooks for the posts linked in Section 1, imported on 2 September 2026. The X supplement used the query conversation_id:2085363072302993719 is:reply between 6 August 2026 13:51 UTC and 2 September 2026, with an approved cap of 2,500. Retrieval was interrupted once and resumed from saved pagination; it ended on an empty page at 2,260 distinct replies. 425 matched existing X rows by identifier and were not reimported; 12 original rows not returned by the search were retained. Exhaustion means the query returned no more results at that time. It does not prove that deleted, private or unindexed replies were recovered.
A source-row audit on 12 October 2026 resolved an apparent date conflict in the original X workbook. Its row dated 15 July is a different, older post; the initiating post is the next row, dated 6 August at 13:51:15 UTC, which matches the record retrieved from X. The older post remains in the frozen DDI dataset but is not a reply to the video. Timestamps for the original exports, absent from the analysis database, were recovered from the sealed workbooks after this analysis was frozen, so timing here is analysed for the supplementary X replies only.
Screening
Screening removed comments below 20 characters or three words (2,705), emoji-only comments (224), unresolved language (46), unsupported language (1) and stimulus material (1). Minimum-length rules under-represent brief approval, disapproval and slogans. No sentiment, relevance or political-position filter was applied. The enlarged cohort cannot be described as wholly verified English-only.
Instrument and composite
The guiding documents are the DDI Methodology Book and the DDI Methodology and Rubric Guide, Second Edition: Comment Edition (September 2026), from GCRD with ReHuman Lab and Sensika Technologies. Only the target comment is scored; parent comments and the post supply context. Popularity, platform, agreement with the political message and the truth of claims must not mechanically determine the core indicators. The composite treats an ordinal rubric additively; this is the instrument’s descriptive convention, not proof of equal psychological distances between anchors.
ReHuman configuration
The primary runs used the ReHuman 0.1.1 route, backed by a Gemini 3 Flash Preview foundation model. Its adaptation to the DDI is prompt-based: the rubric, the 0–4 anchors, context rules and up to five human-reviewed examples loaded for in-context calibration. There is no evidence of study-specific fine-tuning. Temperature and seed were not fixed, and historical prompts were not snapshotted, so individual generations cannot be replayed exactly, although every reported statistic can be recomputed from the released scores.
Agreement
Twenty stored comments are human-reviewed, and 19 have a paired ReHuman output. Exact agreement is 293 of 370 eligible indicator pairs (79.19%); within-one agreement is 93.78%; mean absolute error is 0.286. Using the earliest rather than the latest saved ReHuman output gives 78.38%. Per-indicator agreement ranges from 55.56% (interpersonal trust) to 94.74% (de-escalation). The 370 pairs are clustered in 19 comments; 17 of those comments are earlier pilots and none is from the supplementary X wave. References may derive from model-assisted adjudication, and some reviewed comments may have served as calibration examples. The result supports reporting observed same-study agreement. It does not establish independent out-of-sample accuracy.
An older cross-study benchmark, run on 3 September 2026 with an earlier rubric on Pakistan ground-truth posts, recorded an AI-to-adjudicated ordinal alpha of 0.094 on validation and failed its model-release gate; the human-to-human alpha on the same material was 0.932. That benchmark concerns different material and an older rubric. It is disclosed as historical context and is not Part 2’s agreement statistic. Equally, Part 2’s 79% must not be rewritten as an alpha.
Project-specific agreement, indicator by indicator
Exact agreement between ReHuman and the human-reviewed reference scores on each indicator (18 or 19 reviewed comments per indicator). Pooled: 293 of 370 pairs, 79.19%.
Sensitivity
| Analysis | Comments | Mean | Risk band |
|---|---|---|---|
| Primary: uniform bulk ReHuman | 12,139 | 14.69 | 79.73% |
| Excluding human-reviewed rows | 12,137 | 14.69 | 79.73% |
| All four dimensions active | 12,111 | 14.66 | 79.81% |
| Excluding historical campaign-flagged rows | 11,843 | 14.65 | 79.68% |
| All stored numeric scores (mixed provenance) | 12,166 | 14.69 | 79.73% |
| Original export cohort | 10,559 | 15.17 | 78.77% |
| Supplementary X cohort only | 1,580 | 11.45 | 86.20% |
| Equal platform weighting | 12,139 | 16.88 | 74.39% |
The pooled result remains low under every check. Historical similarity flags were not recomputed on the new X rows, so the repetition checks cover recorded flags only; repeated text is not evidence of bots or coordination.
The further lenses
Reception profiles. k-means clustering (k = 5, 20 initialisations, fixed seed) of the twenty raw indicator scores for the 12,111 comments with all four dimensions active. Solutions from three to eight clusters were compared; five was chosen for interpretability (silhouette 0.26). Profile labels describe indicator patterns and were checked against random samples of comments.
Themes. Twelve multi-label keyword rules applied to comment text with account handles removed. They are a reading aid; dictionary methods miss paraphrase and irony.
Uses of hope. Ordered pattern rules (redefinition, denial, quotation or mockery, wish) applied to comments containing the word. Conditions identified in redefinitions use a second keyword set.
Threads. Replies were linked to a parent comment by exact text match where the parent appears once in the scored cohort (1,949 of 2,187 replies).
Limits
- Capture is non-probability. Deleted, private and filtered comments are unknown, and short reactions are excluded. Findings describe the captured corpus, not the British public or each platform’s audience.
- Platform, source post, audience, interface and capture method vary together, so platform differences cannot be isolated as platform effects. Only X received supplementary retrieval.
- Comments are not people. Authors may post several times, and threads are dependent. No significance tests or confidence intervals are reported, because a conventional margin of error would mislead on a dependent, non-probability sample.
- The study does not establish whether the video generated hope, changed opinion or voting intention, attracted support or opposition, or performed better or worse than the July appeal.
Quotations
Quoted comments are public and reproduced as written, apart from the removal of account handles, names and links, light trimming marked […], and masking of profanity. Each carries its platform, its DDI composite and its reception profile.
Data
A text-free analysis package accompanies this report: pseudonymous row-level scores, all summary tables, the codebook, provenance records, checksums and a script that recomputes the headline statistics without database access. Comment text, author identifiers and exact timestamps are not included.
About this analysis
Author. Jacob Udo-Udo Jacob is the Founding Executive Director of the Global Centre for Rehumanising Democracy (GCRD), a think-act-teach centre working on democratic renewal, AI governance and civic dialogue.
Independence. This is an independent analysis, prepared without commission from any party, campaign or government body.
Acknowledgements. The DDI Comment Edition was developed by GCRD with ReHuman Lab and Sensika Technologies. We thank the university partners whose triple-coding work underpins the wider DDI programme.
The Democracy Discourse Index is a longitudinal, human-validated and AI-assisted system for measuring democratic health through the quality of public discourse. More at gcrd.org.uk/ddi.
Read the whole study
- ReportWhat Hope Means to the Public You are hereThe analytical report: findings, interpretation, and implications for policy and leadership.
- Interactive mapThe Discourse MapEvery scored comment on the map, with a guided reading, a comparison tool and seminar exercises.
- Lexical appendixThe Vocabulary of the ReplyThe words the public used, by platform, reception profile and DDI band.
References
- GCRD (2026). Let’s Bring Back Hope, Part 2: What Hope Means to the Public. GCRD Insights, October 2026.
- GCRD (2026). The Discourse Map: interactive companion to “Let’s Bring Back Hope, Part 2”. GCRD Insights, October 2026.
- GCRD (2026). The Vocabulary of the Reply: lexical appendix to “Let’s Bring Back Hope, Part 2”. GCRD Insights, October 2026.
- GCRD (2026). “Let’s Bring Back Hope” in 48 Hours: the public reception to PM Andy Burnham’s appeal to hope. GCRD Insights, July 2026.
- GCRD (2026). “Let’s Bring Back Hope”: an interactive discourse map (Part 1). GCRD Insights, July 2026.
- GCRD (2026). The vocabulary of the reply: lexical appendix (Part 1). GCRD Insights, July 2026.
- GCRD (2025). A Framework for Understanding AI-Induced Fracture and Authentic Leadership Restoration: A Five-Tier Trust Architecture. GCRD Insights.
- GCRD, ReHuman Lab and Sensika Technologies (2026). DDI Methodology and Rubric Guide, Second Edition: Comment Edition. September 2026.
- GCRD (2026). DDI Methodology Book, Second Edition: Comment Edition. September 2026.
- Jahoda, M., Lazarsfeld, P. F. and Zeisel, H. (1933). Die Arbeitslosen von Marienthal. Leipzig: Hirzel. English edition (1971): Marienthal: The Sociography of an Unemployed Community. Chicago: Aldine-Atherton.
- McLuhan, M. (1964). Understanding Media: The Extensions of Man. New York: McGraw-Hill.
- Sherif, M. and Hovland, C. I. (1961). Social Judgment: Assimilation and Contrast Effects in Communication and Attitude Change. New Haven: Yale University Press.
- Williams, R. (1989). “The Politics of Nuclear Disarmament”. In R. Gable (ed.), Resources of Hope: Culture, Democracy, Socialism, pp. 189–209. London: Verso. First published in New Left Review, 124, November–December 1980.
- The Rule of Saint Benedict, Prologue, verse 1.
Suggested citation. Jacob, J. U. (2026). Let’s Bring Back Hope, Part 2: What Hope Means to the Public. Global Centre for Rehumanising Democracy, October 2026.
Global Centre for Rehumanising Democracy · gcrd.org.uk · info@gcrd.org.uk
