Revolutionizing Blockchain Consensus: Towards Deliberative and
Unanimous Agreement
Syed Badruddoja
1
, Ram Dantu
2
, Mark Dockendorf
2
, Abiola Salau
2
and Kritagya Upadhyay
3
1
Dept. of Computer Science, California State University, Sacramento, 6000 J Street, Sacramento, California, 95819, U.S.A.
2
Dept. of Computer Science, University of North Texas, 3940 N. Elm Street, Denton, Texas, 76207, U.S.A.
3
Dept. of Computer Science, Middle Tennesse State University, 1301 E Main St, Murfreesboro, TN 37132, U.S.A.
Keywords:
Blockchain, Consensus Protocol, Deliberative Consensus, Algorithm, Artificial Intelligence.
Abstract:
Consensus algorithms require a majority of nodes in a distributed system to agree on a single value. Blockchain
systems commission these consensus algorithms to ensure security and trust in decentralized applications.
However, current consensus algorithms do not address the requirements of high-stake applications that demand
unanimous consensus with deliberation. For instance, a trial case at a court requires unanimous consensus to
decide the fate of a criminal. With limited agreement structure and no deliberation, the current consensus
protocol cannot handle the consensus problem. Our research determines the requirements of a deliberative
unanimous consensus model for high-stake applications. Moreover, we propose a family of consensus models
that agree on the answer’s correctness and the methods used to reach it.
1 INTRODUCTION
Shrinking Trust in Real-world Consensus: The
trustworthiness of judicial systems is challenged by
bias and manipulation, which have adverse effects on
society (K.Lin, 2023). According to a recent survey
by the Pew Research Center, less than half of Amer-
icans (44%) currently hold a favorable view of the
court. At the same time, a slim majority (54%) har-
bor an unfavorable opinion. Over the last two years,
the court’s favorable rating has plummeted by 26 per-
centage points. Moreover, according to the monthly
survey conducted by the NJC (National Judicial Col-
lege), most judges hold the belief that systemic racism
exists within the criminal justice system of the United
States(Firth, 2020). In the scientific fields, erroneous
scientific consensus can arise unexpectedly without
any obvious vested interests (Socol et al., 2019) due
to various reasons. The varied interests can intro-
duce biases, ultimately shaping the consensus. Such
cases suffer from decisions made with low to no trust.
(Abraham et al., 2023; McKenzie et al., 2022).
Blockchain - Not a Deliberative Consensus:
Blockchain consensus mechanisms offer to establish
agreement and trust within decentralized networks,
eliminating reliance on a central authority (Lin and
Liao, 2017). This decentralization guarantees that
transactions are validated and recorded by a dis-
Discrete Decision
No deliberation - 50/50 chance
of 100% consensus
No harmony in agreements
due to lack of logical
deliberation
Committee Formation
and Validation in existing
Consensus Mechanisms
No Committee in PoW
Committee formed on basis
of stake or money in PoS
Sybil Attacks
Sybil Nodes
Good Nodes
Only needs 51% to overturn
Absence of
Deliberation
Figure 1: Current consensus protocol limitations and prob-
lems that cannot solve high-stake consensus problem and
fails to deliberate among participants.
tributed network of nodes, enhancing security, re-
silience, and resistance to amendments. However, the
consensus mechanism of blockchain is fairly simple
and does not reflect a true deliberative consensus. The
agreement in computer systems is limited to match-
ing the same outcome irrespective of any argument
or arbitration (Xiao et al., 2019). Figure 1 shows the
problems with current consensus protocols.
Unreliable Consensus Protocols: Consensus prob-
lems in distributed systems have to agree on a spe-
cific issue even in the presence of faulty/malicious
agents(Kshemkalyani and Singhal, 2011). (Lamport
and Fischer, 1982) in 1982 described the Byzantine
general’s problem. This scenario depicts a commu-
nication problem among a certain number of gener-
786
Badruddoja, S., Dantu, R., Dockendorf, M., Salau, A. and Upadhyay, K.
Revolutionizing Blockchain Consensus: Towards Deliberative and Unanimous Agreement.
DOI: 10.5220/0012813400003767
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 21st International Conference on Security and Cryptography (SECRYPT 2024), pages 786-791
ISBN: 978-989-758-709-2; ISSN: 2184-7711
Proceedings Copyright © 2024 by SCITEPRESS – Science and Technology Publications, Lda.

Table 1: Existing Consensus Protocols and the limited Consensus Agreement Percentage.
Protocol Name Consensus Type Committee/Leader
Selection
Adversary
Tolerance
Network
Synchrony
Snow White (Da-
ian et al., 2019;
Buterin et al.,
2020)
PoS (Sleepy
Model)(Kiayias et al.,
2017)
MPC < 51% stakes Asynchronous
Ouroboros (David
et al., 2018)
PoS MPC (PVSS) < 51% stakes Partial Syn-
chrony
Ouroboros
Praos(Gilad et al.,
2017)
PoS VRF < 51% stakes Partial Syn-
chrony
Algorand (Bentov
et al., 2016)
PoS-BFT, PoS-based
for block proposing and
Byzantine agreement for
block finalization
VRF (Cryptographic
sortition)
< 33%
weighted
users
Asynchronous
periods
between
synchronous
periods
Casper FFG
(Nakamoto,
2008)
Light-weight PoS layer
over Ethereum PoW
PoW-based leader se-
lection BFT for block
checkpoint justification
< 51% valida-
tors
Partial Syn-
chrony
Bitcoin (Kwon,
2014)
PoW Based on hash < 51% com-
puting power
Asynchronous
Tendermint
(Buchman, 2016)
PoS-BFT Round-robin < 33% voting
power
Partial Syn-
chrony
als located in different locations, which is a challenge
of validating messages sent from one to the other.
However, they lack the basis for a consensus in cer-
tain real-life situations or systems that require or even
mandate true deliberation, such as a jury consensus
problem. To add more to it, the traditional consensus
mechanisms fail to address the high-stake consensus
demands of real-world situations. In the PoW system
(Lin and Liao, 2017), the block creation is confirmed
by verifying the nonce of the block that produces an
equivalent or lesser hash value than the target value
proposed by the prover. In PoS systems (Chen and
Micali, 2019), the block proposers propose a block
with a random hash value in their block proposal.
However, there is hardly any relation between the
consensus protocol and real-world high-stakes con-
sent problems where a group of people must decide
on the fate of a crime committed by a criminal.
Limited Agreement Quorum: Various blockchain
consensus protocols follow different agreement per-
centages concerning proving the majority agreement
of the agreement problem. PoW, PoS, consider a 51%
majority agreement, whereas Pure PoS with practi-
cal byzantine fault tolerance failure method follows a
two-third majority validation rule(Gilad et al., 2017).
While these types of agreements are favorable from
a computer failure perspective, they do not guarantee
that the decision can be trusted for any real-life con-
sensus problem. Table 1 summarizes some of the ma-
jor consensus protocols that fail to address unanimity
in consensus.
We investigate the requirements for creating a de-
liberative consensus for high-stake applications such
as ”Jury Trial Decisions” and propose a high-level so-
lution for the same.
• We outline the requirements of unanimous con-
sensus for blockchain applications that will sup-
port critical high-stake applications.
• We propose an architecture for consensus in three
formats: Unanimous Consensus with Weak De-
liberation (UCWD), Unanimous Consensus with
Strong Deliberation (UCSD), and Unanimous
Consensus with Algorithmic Election (UCAE).
• For each consensus proposal, we proposed an
agreement method to achieve unanimous consen-
sus under specific circumstances using algorith-
mic proofs.
2 LITERATURE REVIEW
Biased Consensus in Real-World: The bias in
courtroom decisions has plagued the judicial systems
and affected the social welfare of the people over
Revolutionizing Blockchain Consensus: Towards Deliberative and Unanimous Agreement
787

the last few years. According to the National Judi-
cial College (University of Nevada, Reno) ’s monthly
survey of its alumni published in 2020, the major-
ity of judges believe that racism is systemic in the
United States criminal justice system (Firth, 2020).
In the survey, 65 percent of the 634 judges believe
that systemic racism exists in the criminal justice sys-
tem. More than 200 judges left comments with their
votes, and the consensus among the majority was that
racism is mostly of the implicit or unconscious kind.
Another example of biased consensus is seen in Sci-
entific research. Socol et al. mentioned that, human
scientists are susceptible to bias, influenced by politi-
cal and economic interests (Socol et al., 2019).
Agent-based Consensus: Multi-agent consensus
system addresses the consensus problem through de-
liberations. Hadfi et al. (Hadfi and Ito, 2022) pro-
pose the development of autonomous and intelligent
conversational agents that can augment the delibera-
tive capacities of citizens in social media. The au-
thors proposed an approach that quantifies delibera-
tion in online argumentative discussions. Moreover,
Zhang et al. developed a web tool that enables peo-
ple to create and evaluate Machine Learning models
in order to examine the strengths and shortcomings
of past decision-making and deliberate on how to im-
prove future decisions. The authors applied the tool to
improve graduate school admission decisions (Zhang
et al., 2023). However, these developments assume
that data input for unbiased decisions is pristine and
does not guarantee immutability.
To the best of our knowledge, no work has been
found to address high-stake, deliberative digital con-
sensus that is tamper-proof and handles real-world ap-
plications. This paper proposes a novel approach to
true consensus by recommending requirements and
introducing an algorithmic deliberation approach to
solve the consensus problems discussed so far.
3 REQUIREMENTS FOR
UNANIMITY
Valid Belief and Selection of Participants: Cur-
rently, in blockchain’s consensus protocol mecha-
nism, the belief in the designated validators and block
proposals is solely based on money and computa-
tional power or both without any deliberations. Some
participants in the consensus mechanism might have
less stake or power but more knowledge, facts, and
experience, which could be more valuable in coming
to a proper agreement and decision-making process.
Although the current systems only allow that partic-
ipant to be the validator or decision-maker who has
the highest stake or the highest computational power,
there should be a provision where the participants for
the consensus would not be ruled out just based on
their comparatively fewer stakes and computational
power than others.
Rake and Quantify Trust: The systems that in-
volve multiple peers, entities, and nodes are anony-
mous and lack accountability in blockchain. There-
fore, this leaves the door open for the entities and
nodes with malicious intent that can corrupt the
decision-making process and put unanimity on hold.
Nevertheless, once the trust is gathered from the set
of prospective entities in the form of their knowledge,
experience, stakes, and computational power through
different series of exhaustive deliberations, each en-
tity or peer in the system can be assigned a unique
trust value based on the gathered data.
Logic on Deliberations to Arrive at a Result:
There should be a formal model and logic in which
the involved entities and nodes in the consensus han-
dle and perform their deliberation. For instance, the
Quaker-based model can be used for the deliberation
process.
Broadcast and Synchronize the Deliberation: All
the arguments, proofs, and theorems used for the de-
liberation should be shared and synchronized across
multiple sites, institutions, or geographies and made
accessible to all involved entities for the decision-
making process.
Detect and Prevent Gang-Up Coalition: In spite
of the fact that unanimity or 100% consensus is the
major goal of the paper to establish trust, it does not
mean that unanimity by force or unanimity by coali-
tion is left undetected and overlooked as it defeats the
whole purpose of 100% consensus via deliberation.
Hence, the detection and prevention of the gang-up
alliance is a significant requirement.
Forward-Looking Consensus Model: With quan-
tum and AI architectures rising to join classical com-
puters, there is a need to allow multiple fundamentally
different architectures to form consensus together.
Some of these will be predefined/scripted programs;
others will be ”smart” entities that are capable of ma-
chine learning. Figure 2 shows a paradigm shift of
traditional real-world consensus to true deliberative
SECRYPT 2024 - 21st International Conference on Security and Cryptography
788

AI-Based
Deliberation
Blockchain
Consensus
Smart
Legal Contract
Group
Discussion
Government
Policies
Corporate
Agreeements
Judicial
Convictions
Negotiate/
Arbitrate
Mediate/
Adjudicate
High-Stake
Consensus
Application
Next-gen
Methods of
Resolution
Traditional
Methods of
Resolution
Figure 2: Paradigm shift from traditional biased real-world
consensus to distributed deliberative consensus applications
for high-stake applications.
digital consensus for high-stake applications that will
reduce bias and increase trust in the system.
4 ARCHITECTURE
We describe three variations of unanimous consensus
(UC): Unanimous Consensus with Strong delibera-
tion (UCSD), Unanimous Consensus with Weak De-
liberation (UCWD), and Unanimous Consensus with
Algorithm Election (UCAE). While all of these exam-
ples require all committee members (deliberators) to
be honest, it only requires that at least one is provably
correct to form a consensus. Furthermore, UCWD
and UCSD do not require deliberators to have the
same architecture or execute the same instructions to
reach a consensus. This allows these two forms of
UC to be used in systems with a mix of classical
and quantum machines.Figure 3 is our proposed ar-
chitecture. We develop our consensus protocol with
4 phases. In phase 1, decentralized application users
request transactions on the blockchain. In phase 2,
the transactions are dequeued and sent to phase 3 for
deliberation in different consensus engines. Phase 3
completes consensus and sends the results for block
creation in phase 4. One example of a formal method
of deliberation is mathematical proving, as mathe-
matics serves as a ground truth for all modern sci-
ence. With automated theorem proving, it is possi-
ble to prove/disprove mathematical statements within
stated bounds.
Validation: The correctness of an algorithm is so-
lidified by a proof. Before an algorithm is accepted,
its proof is checked to ensure it is correct by ensuring
Figure 3: Ideal architecture of unanimous consensus where
various transactions from various DApps in the network are
queued and then dequeued for extensive deliberation in the
consensus engine.
every logical step follows from the previous step(s)
within the stated constraints. Secondly, the evidence
input to the algorithm must fit the constraints. Fi-
nally, the algorithm must result in the desired output.
For instance, if the network graph were to have nega-
tive edges, Dijkstra’s algorithm to compute the short-
est path would be invalid due to data incompatibility
with the constraints of the proof for Dijkstra’s algo-
rithm. Similarly, in a jury consensus, a deliberator can
declare that they cannot solve the problem. If a de-
liberator declares they cannot solve the problem, this
counts as an abstention. A verdict cannot be reached
if any deliberators abstain at the end of the round (this
means that another round of deliberations is needed).
Suppose a deliberator cannot solve a problem initially
(i.e., lacks a necessary algorithm). In that case, they
can adopt an algorithm that was offered by one of the
other deliberators and proven correct.
Analysis: This phase varies by deliberation type.
Weak and strong deliberation consists only of running
the selected algorithm to find the answer. In unani-
mous consensus with algorithm election, a validated
algorithm is elected first, and then each deliberator
runs the selected algorithm. UCAE should use only
deterministic algorithms.
Outcome: There will be several rounds of deliber-
ation for consensus in a deliberative blockchain con-
sensus system. After all rounds have been completed,
there are two possible states: verdict or hung for a
jury consensus case. If all deliberators cannot agree
that the output is correct or all deliberators agree that
they cannot solve the problem, then the deliberators
are hung. This is equivalent to other consensus algo-
Revolutionizing Blockchain Consensus: Towards Deliberative and Unanimous Agreement
789

Problems, Inputs,
& Constraints
A C DB
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Result
(1)
(2)
(3)
Problems, Inputs,
& Constraints
A C DB
All-to-All Exchange
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Run
Chosen
Algorithm
Result
Validate
Peers
Validate
Peers
Validate
Peers
Validate
Peers
(1)
(2)
(3)
(4)
(5)
Problems, Inputs,
& Constraints
A C DB
All-to-All Exchange
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Algorithm
Proof
Run
Elected
Algorithm
Run
Elected
Algorithm
Run
Elected
Algorithm
Run
Elected
Algorithm
Result
Validate
Peers
Validate
Peers
Validate
Peers
Validate
Peers
Elect Algorithm
(1)
(2)
(3)
(4)
(5)
(6)
Figure 4: From the left, the first one is the UCWD type consensus algorithm, in the middle is the UCSD consensus algorithm,
and on the right is the UCAE consensus algorithm.
rithms failing to form a consensus (e.g., for a block in
a blockchain network) and should be expected to have
similar consequences. When a verdict is reached in
weak deliberation, it means that all deliberators agree
that the provided output is correct. When a verdict
is reached with strong deliberation, all deliberators
agree that each deliberator’s method to obtain the out-
put is correct (proven and peer-reviewed) and the out-
put itself is correct. A verdict in algorithm election
means an algorithm was selected from the proposed
algorithms, which had correct proofs, all deliberators
ran the selected algorithm, and the outcome was cor-
rect according to the deliberators. Figure 4 shows
three types of blockchain-based unanimous consen-
sus discussed in this section.
4.1 Unanimous Consensus with Strong
Deliberation (UCSD)
To reach a strongly deliberated unanimous consensus,
all deliberators must agree on the answer and agree
that the algorithms used by their peers to reach their
conclusions are correct. This is done by exchanging
algorithms and their corresponding proofs early in the
consensus round and validating their peers’ proofs.
Even if all deliberators agree, another round of con-
sensus may be needed as it is possible to reach the
correct output with incorrect logic. If multiple out-
puts are produced by deliberators and multiple out-
puts are not desired, which may be possible with some
problems (e.g., shortest path, minimum spanning tree,
etc.), answers can be (but are not required to be, as the
algorithm was proven correct earlier) checked to en-
sure they are correct (e.g., all minimum spanning trees
for a graph should have the same total edge weight),
then a vote is held to elect an answer among the cor-
rect outputs. The user can set heuristics for select-
ing the desired correct answer (e.g., majority, lowest
hash, etc.).
4.2 Unanimous Consensus with Weak
Deliberation (UCWD)
Weak deliberation is an optimization that skips the
need to check algorithm proofs if all deliberators
reach the same initial conclusion. This optimiza-
tion relies upon at least one deliberator being cor-
rect. In realistic systems where strong deliberation
has been resolved reliably on the first round with sim-
ilar problems, UCWD can provide large speedups.
These speedups will be more readily noticeable when
the time complexity of proving the algorithm domi-
nates the time complexity of running the algorithm:
O(algorithms) ¡ O(validation). If UCWD fails to
reach a consensus in the first round, it can fall back
to strong deliberation (UCSD). This can be consid-
ered ”only deliberate if we need to”. Weak delibera-
tion should not be used when the problem likely has
multiple correct answers.
4.3 Unanimous Consensus with
Algorithm Election (UCAE)
Beyond strong deliberation, the algorithm must be the
same for all machines. Unlike the previous two mech-
anisms, this one will require either a virtual environ-
ment or similar hardware. Unlike strong and weak
deliberation, which allow the mixing of quantum and
classical computers, algorithm election requires that
all hardware perform the selected algorithm. This
SECRYPT 2024 - 21st International Conference on Security and Cryptography
790

form of unanimous consensus is the closest to exist-
ing consensus algorithms: it is provable that perform-
ing the same operations in the same environment with
the same input values results in the same output (e.g.,
executing a smart contract should yield the same re-
sult for all machines – this allows blockchains with
embedded programs to form consensus). Despite its
shortcomings, algorithm election excels in one area:
forming a unanimous consensus around a single out-
put when a problem may have multiple correct an-
swers.
5 CONCLUSION
Blockchain consensus protocols evolved to guarantee
the security of applications with a consensus-based
agreement between multiple parties in the network.
However, real-world high-stakes applications such as
trials at a jury cannot depend on blockchain for the
output of consensus agreements as it suffers from in-
complete agreement percentages and non-deliberative
decisions. Hence, the consensus mechanisms of
blockchain suffer from low trust. To overcome this
difficulty, we proposed a deliberative consensus pro-
tocol with unanimous agreement. First, we described
the requirements of unanimous consensus. Secondly,
we proposed solutions under UCSD, UCWD, and
UCAE to achieve a consensus on blockchain. It takes
more time and effort to reach a conclusion when the
decision is made by deliberation and unanimity. In
our future work, we will explore and solidify the the-
oretical work for unanimous consensus.
REFERENCES
Abraham, J. et al. (2023). Science, politics and the pharma-
ceutical industry: Controversy and bias in drug regu-
lation. Routledge.
Bentov, I., Pass, R., and Shi, E. (2016). Snow white: Prov-
ably secure proofs of stake. iacr cryptol. eprint arch.
Buchman, E. (2016). Tendermint: Byzantine fault tolerance
in the age of blockchains. PhD thesis, University of
Guelph.
Buterin, V., Reijsbergen, D., Leonardos, S., and Piliouras,
G. (2020). Incentives in ethereum’s hybrid casper pro-
tocol. International Journal of Network Management,
30(5):e2098.
Chen, J. and Micali, S. (2019). Algorand: A secure and ef-
ficient distributed ledger. Theoretical Computer Sci-
ence, 777:155–183.
Daian, P., Pass, R., and Shi, E. (2019). Snow white: Ro-
bustly reconfigurable consensus and applications to
provably secure proof of stake. In Financial Cryp-
tography and Data Security: 23rd International Con-
ference, FC 2019, Frigate Bay, St. Kitts and Nevis,
February 18–22, 2019, Revised Selected Papers 23,
pages 23–41. Springer.
David, B., Ga
ˇ
zi, P., Kiayias, A., and Russell, A.
(2018). Ouroboros praos: An adaptively-secure, semi-
synchronous proof-of-stake blockchain. In Advances
in Cryptology–EUROCRYPT 2018: 37th Annual In-
ternational Conference on the Theory and Applica-
tions of Cryptographic Techniques, Tel Aviv, Israel,
April 29-May 3, 2018 Proceedings, Part II 37, pages
66–98. Springer.
Firth, A. (2020). Most judges believe the criminal justice
system suffers from racism. National Judicial COl-
lege, University of Nevada, Reno.
Gilad, Y., Hemo, R., Micali, S., Vlachos, G., and Zeldovich,
N. (2017). Algorand: Scaling byzantine agreements
for cryptocurrencies. In Proceedings of the 26th sym-
posium on operating systems principles, pages 51–68.
Hadfi, R. and Ito, T. (2022). Augmented democratic delib-
eration: Can conversational agents boost deliberation
in social media? In Proceedings of the 21st Interna-
tional Conference on Autonomous Agents and Multia-
gent Systems, pages 1794–1798.
Kiayias, A., Russell, A., David, B., and Oliynykov, R.
(2017). Ouroboros: A provably secure proof-of-stake
blockchain protocol. In Annual international cryptol-
ogy conference, pages 357–388. Springer.
K.Lin, C, D. (2023). Favorable views of supreme court fall
to historic low. Pew Research Center.
Kshemkalyani, A. D. and Singhal, M. (2011). Dis-
tributed computing: principles, algorithms, and sys-
tems. Cambridge University Press.
Kwon, J. (2014). Tendermint: Consensus without mining.
Draft v. 0.6, fall, 1(11):1–11.
Lamport, L. and Fischer, M. (1982). Byzantine generals and
transaction commit protocols.
Lin, I.-C. and Liao, T.-C. (2017). A survey of blockchain
security issues and challenges. Int. J. Netw. Secur.,
19(5):653–659.
McKenzie, N. D., Liu, R., Chiu, A. V., Chavez-MacGregor,
M., Frohlich, D., Ahmad, S., and Hendricks, C. B.
(2022). Exploring bias in scientific peer review: an
asco initiative. JCO oncology practice, 18(12):791–
799.
Nakamoto, S. (2008). Bitcoin whitepaper. URL:
https://bitcoin. org/bitcoin. pdf-(: 17.07. 2019), 9:15.
Socol, Y., Shaki, Y. Y., and Yanovskiy, M. (2019). Interests,
bias, and consensus in science and regulation. Dose-
Response, 17(2):1559325819853669.
Xiao, Y., Zhang, N., Li, J., Lou, W., and Hou, Y. T.
(2019). Distributed consensus protocols and algo-
rithms. Blockchain for Distributed Systems Security,
25:40.
Zhang, A., Walker, O., Nguyen, K., Dai, J., Chen, A., and
Lee, M. K. (2023). Deliberating with ai: Improv-
ing decision-making for the future through participa-
tory ai design and stakeholder deliberation. Proceed-
ings of the ACM on Human-Computer Interaction,
7(CSCW1):1–32.
Revolutionizing Blockchain Consensus: Towards Deliberative and Unanimous Agreement
791